JPH1086081A - Walking posture generation device for leg type moving robot - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide appropriate steps and rotational angles by providing a walking posture generation means which performs approximate calculation of requirements on the walking attitude based on a selected standard walking posture and generates an approximate walking posture for satisfying the requirements on the walking posture. SOLUTION: A standard walking posture data group which is prepared on off-line is stored in a ROM beforehand (S10). Secondly, an end upper body position and speed of each standard walking posture are found so as to be stored (S12). A timer value (t) is set to 0 (S14) so that interruption of the timer is waited (S18). Thirdly, a changing point of the walking posture is judged and when it is affirmative, the timer is reset (S20) and a requirement value to the walking posture is read (S24). Fourthly, generation processing of the walking attitude is implemented (S26). Fifthly, the walking posture generation parameter is substituted (S28) for a target walking posture parameter and an instantaneous value of the target walking posture is calculated (S30). Sixthly, the time (t) is renewed (S32) by Δt. Both-leg compliance is provided based on the walking posture generated thereby, a target joint angle is found, and the robot is driven and controlled for the target value.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a gait generator for a legged mobile robot, and more particularly to a legged mobile robot capable of generating its gait freely and in real time.

[0002]

2. Description of the Related Art In a conventional legged mobile robot, various gaits such as straight-ahead travel and direction change are designed off-line in advance, as disclosed in Japanese Patent Application Laid-Open No. 5-285868. It has been proposed to store the information in a memory of a control microcomputer mounted on a robot and output and execute the data in a suitable order during walking.

[0003] Further, in the technique described in Japanese Patent Application Laid-Open No. Sho 62-97006 (Japanese Patent Publication No. 5-62363), walking pattern data generated off-line is stored in a memory in advance, and is output at the time of walking. It is proposed to subdivide the step time by temporally interpolating between the time series data stored according to.

[0004]

However, in the above-mentioned prior art, since the gait data is predetermined, the gait can be freely and real-time generated, such as realizing an arbitrary stride or turning angle. I couldn't do that.

A first object of the present invention is to solve the above-mentioned disadvantages of the prior art, and to freely and in real time generate a gait including a floor reaction force in a legged mobile robot and to set an arbitrary stride, turning angle and the like. It is an object of the present invention to provide a gait generation device for a legged mobile robot which can be realized.

A second object of the present invention is to solve the above-described disadvantages of the prior art, to generate a gait freely and in real time in a legged mobile robot, and to control various parts of the robot at the boundaries between the generated gaits. It is an object of the present invention to provide a gait generating device for a legged mobile robot in which the displacement and the speed of the robot are continuous.

A third object of the present invention is to solve the above-mentioned disadvantages of the prior art, and in a legged mobile robot, to drive and control the legged mobile robot based on a gait generated in real time. An object of the present invention is to provide a gait generating device for a robot.

[0008]

In order to achieve the first and second objects, for example, in claim 1, at least an upper body and a plurality of bodies connected to the upper body via joints. In a gait generating device for a legged mobile robot including a leg link, a standard gait storing a plurality of types of standard gaits each comprising a set of parameters including a parameter relating to a floor reaction force for at least one gait. Gait storage means, gait request means for making a request relating to a gait, and selecting one or more of the standard gaits in response to the request relating to the gait, and performing an approximate operation based on the selected standard gait. Thereby, the gait is provided with gait generating means for generating an approximate gait satisfying the requirement regarding the gait.

According to a second aspect of the present invention, the gait generator is configured to obtain the approximate gait by performing a weighted average of the selected gait.

According to a third aspect of the present invention, the gait generator is configured to obtain the approximate gait using the parameter sensitivity of the selected gait.

According to a fourth aspect of the present invention, the gait generating means is configured to calculate a weighted average of the selected gait and obtain the approximate gait using parameter sensitivity of the selected gait. .

According to a fifth aspect of the present invention, the gait generating means calculates a change amount of another parameter per unit change amount of at least one of the standard gaits as a parameter sensitivity of the gait. There is provided storage means for storing, and the approximate gait is obtained using the parameter sensitivity of the stored gait.

According to a sixth aspect of the present invention, the standard gait generating means includes expression means for expressing a relation of a perturbation of another parameter to a parameter relating to a floor reaction force of the standard gait. The gait is configured to correct the approximate gait based on the perturbation relationship and generate a gait that satisfies the requirement for the gait so that at least displacement and velocity are continuous under boundary conditions.

In a preferred embodiment, the perturbation of the other parameters is a perturbation of the horizontal position of the body.

According to the present invention, the relationship is represented by a linear model.

According to a ninth aspect of the present invention, the relationship is represented by an inverted pendulum model.

According to a tenth aspect of the present invention, the relationship is represented by a time-series value.

In another preferred construction, the gait generating means calculates the approximate gait by excluding the parameter when the parameter relating to the floor reaction force is under a predetermined condition.

According to a twelfth aspect of the present invention, the gait generator includes a joint angle calculator for calculating a joint angle of the robot based on the generated gait, and the calculated joint angle. And an articulation control means for driving and controlling the joints of the robot.

According to a thirteenth aspect of the present invention, the joint angle determining means detects a state quantity of the robot,
Correction means for correcting the position and orientation of the robot based on the detected state quantity, wherein the joint angle calculation means calculates the joint angle of the robot so as to obtain the corrected position and attitude. did.

[0021]

The gait including the floor reaction force can be freely and in real time generated to realize an arbitrary stride, turning angle, and the like. Further, the displacement and the speed of each part of the robot can be made continuous at the boundary between the generated gaits.
Further, it is possible to drive and control the legged mobile robot based on the gait generated in real time.

[0022]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS A gait generator for a legged mobile robot according to the present invention will be described below with reference to the accompanying drawings.
A bipedal walking robot is taken as an example of the legged mobile robot.

FIG. 1 is an explanatory diagram of the entire device including the bipedal walking robot 1.

As shown in the figure, the bipedal walking robot 1 has six joints on each of the left and right leg links 2 (for convenience of understanding, each joint is shown by an electric motor that drives it). The six joints are, in order from the top, joints 10R and 10L for turning the crotch (waist) (R is the right side and L is the left side).
The same shall apply hereinafter), joints 12R and 12L in the crotch (lumbar) pitch direction (around the Y axis), joints 14R and 14L in the same roll direction (around the X axis), and joints 16R and 1 in the knee pitch direction.
6L, joints 18R and 18L in the pitch direction of the foot, and joints 20R and 20L in the same roll direction.

The feet 22R and 22L are attached to the feet, and an upper body (substrate) 24 is provided at the top. A control unit 26 including a microcomputer described later with reference to FIG. Is stored. In the above, the hip joints (or hip joints) are joints 10R (L) and 12R.
From (L) and 14R (L), the ankle joint is 18R
(L) and 20R (L). The hip joint and the knee joint are connected by thigh links 28R and 28L, and the knee joint and the ankle joint are connected by crus links 30R and 30L.

With the above configuration, the leg link 2 is given six degrees of freedom for each of the left and right feet, and by driving these 6 * 2 = 12 joints at an appropriate angle during walking, A desired movement can be given to the whole, and the user can arbitrarily walk in a three-dimensional space ("*" indicates multiplication in this specification).

The position and speed of the upper body, which will be described later in this specification, are determined at predetermined positions of the upper body 24, specifically, the upper body 2
4 means a representative point such as the position of the center of gravity and its (displacement) speed.

As shown in FIG. 1, a well-known six-axis force sensor 44 is attached below the ankle joint, and three-directional components Fx and F of the force.
By measuring y, Fz and the three-direction components Mx, My, Mz of the moment, the presence or absence of the foot landing or the contact load is detected. An inclination sensor 60 is installed on the body 24 to detect an inclination with respect to the Z axis (vertical direction) and its angular velocity. The electric motor of each joint is provided with a rotary encoder for detecting the amount of rotation.

Further, although not shown in FIG. 1, a joystick 62 is provided at an appropriate position of the bipedal walking robot 1, and a required gait such as turning a robot that is proceeding straight forward from the outside as necessary is required. Is configured to be input.

FIG. 2 is a block diagram showing details of the control unit 26, which is constituted by a microcomputer. The output of the tilt sensor 60 and the like is converted into a digital value by the A / D converter 70, and the output
Via the RAM 74. The output of an encoder disposed adjacent to each electric motor is input to the RAM 74 via the counter 76.

In the control unit, a first unit comprising a CPU,
Second computing devices 80 and 82 are provided, and the first computing device 80 generates a gait freely and in real time based on a standard gait stored in a ROM 84 and a target joint angle as described later. Is calculated and sent to the RAM 74. Further, the second arithmetic unit 82 reads the target value and the detected actual value from the RAM 74, calculates a control value necessary for driving each joint, and calculates each control value via the D / A converter 86 and the servo amplifier. To the electric motor that drives the motor.

FIG. 3 is a block diagram functionally showing the operation of this apparatus, and FIG. 4 is a flow chart (structured flow chart) showing gait mixing or gait generation processing therein. In the specification, gait generation is also referred to as “gait mixing”.

First, the gait generated by this device will be described. An object of the present invention is to provide a device for generating a gait in real time as described above. It is an object of the present invention to freely generate a desired gait required for performing robot posture stabilization control by double leg compliance control proposed in Japanese Patent Application Laid-Open No. Hei 5-305585.

First, the attitude stabilization control proposed above will be briefly described. In this technique, a deviation between a target position and a measured position of a floor reaction force (ZMP) is detected, and the deviation is eliminated. One or both of the legs are driven to stabilize the posture. At the same time, when the robot is about to fall, the actual floor reaction force is shifted by intentionally shifting the target floor reaction force to obtain a posture restoring force.

As described above, in the legged mobile robot, stable walking cannot be realized unless the relationship between the target trajectory and the desired floor reaction force pattern satisfies the dynamic equilibrium condition. Specifically, the dynamic equilibrium condition is, for example, a target floor reaction force center point (an action point on the floor where the second moment of the target floor reaction force distribution becomes zero) and a ZMP (Zero Moment Poi).
nt ... the point where the combined force of inertia and gravity due to motion intersects the floor)
Is to match. If they do not match, when the compliance control is activated, the dynamic balance between the inertial force, the resultant force of gravity, and the floor reaction force is broken, and the vehicle falls down.

In the technique proposed above, it is applied to the recovery of the posture when the robot becomes unstable. However, as is apparent from the above description, in the legged mobile robot, the walking control is used in this way. As the given target value, not only the desired movement pattern but also a desired floor reaction force pattern satisfying the desired movement pattern and the dynamic equilibrium condition is required.

The desired floor reaction force is generally expressed by an action point, a force acting on the point, and a moment of the force.
Since the point of action is good for everywhere, countless expressions are conceivable even with the same desired floor reaction force, but especially when the above-mentioned target floor reaction force center point is used as the point of action to express the desired floor reaction force, the moment of force is Excluding the component perpendicular to the floor, it will be zero.

As described above, the desired floor reaction force center point trajectory satisfying the dynamic equilibrium condition with the desired motion trajectory is ZMP.
This is the same as saying that the desired ZMP trajectory is given instead of the desired floor reaction force action center point trajectory.

Therefore, the above can be paraphrased as "a target ZMP trajectory (a desired floor reaction force pattern) is required as a target value given to the walking control, as well as a target motion trajectory". From such a background, in this specification, a desired gait is defined as follows.

A) A desired gait in a broad sense is a set of a desired motion trajectory during one or more steps and a desired floor reaction force pattern thereof. b) A desired gait in a narrow sense is a set of a desired motion trajectory and a ZMP trajectory thereof during one step. c) In a series of walking, several gaits are connected.

In the following, in order to facilitate understanding,
Unless otherwise stated, the desired gait is used in the narrow sense of the desired gait. More specifically, in this specification, the target gait is used from the beginning of the two-leg support period to the end of the one-leg support period. Needless to say, the two-leg supporting period is a period in which the robot 1 supports its own weight with both the leg links 2, and the one-leg supporting period is a period in which the robot 1 supports one of the leg links 2. The leg (link) on the side that does not support the weight of the robot 1 during the one-leg supporting period is called a free leg.

[0042] The present invention specifically aims at freely and in real time generating the desired gait defined above. If it can be generated freely and in real time, for example, the operator can freely control the robot 1 by remote control, and in automatic control, it can not only move according to a fixed sequence but also perform guidance control and delicate positioning. Becomes

Here, conditions as a desired gait will be described.

The conditions that must be satisfied by the desired gait are roughly classified into the following five. Condition 1) The dynamic equilibrium condition is satisfied. That is,
Z dynamically calculated from the target motion trajectory of the robot 1
The MP trajectory matches the target ZMP trajectory. Condition 2) When the walking plan part and the walking route guidance part (both not shown) of the robot 1 or conditions required by the operator to satisfy the gait, such as the stride length and the turning angle, are satisfied. .

Condition 3) Kinematics (Kinematics) constraints such as foot digging or rubbing on the floor, joint angle not exceeding the movable range, joint velocity not exceeding the limit, etc. . Condition 4) During the one-leg supporting period, ZMP must satisfy the dynamics-restricting conditions such as being within the foot contact surface of the supporting leg and not exceeding the maximum capacity of the driving system.

Condition 5) The boundary condition is satisfied. That is, as a natural consequence of condition 1), at the boundary between gaits, at least a boundary condition that displacement and (displacement) speed of each part are continuous is derived (if discontinuous, infinite Or the ZMP moves to a point farther from the ground contact surface).

Also, the initial state of the (n + 1) th gait is the nth
It must be set to match the end state of the gait (particularly, the position, posture and speed of the upper body with respect to the foot position). At this time, the terminal state of the n-th gait is n +
If the initial state of the first time gait is determined, it may be made to match the initial state of the (n + 1) th gait.

If it is not determined, it is only necessary that the terminal state of the n-th gait falls within a range where the posture can be maintained and long-term walking can be performed. However, as will be described later, it is extremely difficult to determine the range of the terminal state for long-term walking without changing the posture.

In general, the desired gait is generated by a gait generation algorithm including a parameter value or a time-series table, as in the case of a standard gait to be described later. It is nothing but setting the time series table appropriately).

Various gaits are generated by changing the parameter values or the time series table. But,
It is not possible to determine whether or not the created gait satisfies all the gait conditions just by setting the parameter values or the time series table without sufficient consideration.

In particular, even if the range of the terminal body position and speed suitable for long-term walking is known, a body trajectory satisfying the above condition 1) is generated according to the gait generation algorithm based on the ZMP trajectory. In such a case, it is extremely difficult to set parameters related to the ZMP trajectory so that both the body position and the speed at the end of the generated gait fall within the range.

The reason is as follows. Reason 1) Once the upper body is far away from the ZMP, there is a tendency for the upper body to diverge farther away. In order to explain this, the behavior of an inverted pendulum close to the behavior of the robot's upper body will be described as an example.

When the floor projection point of the center of gravity deviates from the fulcrum, the inverted pendulum diverges and falls. However, at this time, the inertial force of the inverted pendulum and the resultant force of gravity act on the fulcrum (ie, Z
MP coincides with the fulcrum), and is balanced with the floor reaction force received from the fulcrum. That is, the dynamic equilibrium condition only represents the relationship between the inertial force, gravity, and floor reaction force of the moving object at that moment.

If the dynamic equilibrium condition is satisfied, it is easy to make an illusion that the long-term walking of the robot is guaranteed, but it has nothing to do with whether or not the posture of the robot is collapsed. Just as the center of gravity of the inverted pendulum moves away from directly above the fulcrum, it tends to move farther away. Just as the center of gravity of the robot moves farther from directly above the ZMP, the divergence tends to move farther away. There is.

Reason 2) Since there are severe restrictions such as that the ZMP must be within the foot contact surface of the support leg during the one leg support period, the acceleration / deceleration pattern of the upper body cannot be set arbitrarily. However, when trying to match the positions, the speeds do not match, and when trying to match the speeds, the positions do not match, making it difficult to match both at the same time. For example, when driving a vehicle, the brake pedal force is limited to a certain narrow range, and the vehicle stops exactly on a target stop line at a predetermined time.

Due to the above problems, it has been impossible in the past to determine the parameters and tables of the desired gait satisfying all the gait conditions in real time. Therefore, free walking has not been realized in the above-described related art.

On the premise of the above, an apparatus according to the present invention will be described with reference to FIG. FIG. 4 is a block diagram functionally showing the operation of the apparatus shown in the flow chart of FIG. 4 for convenience of understanding.

In this apparatus, as shown in the figure, based on a standard gait created off-line and stored in the ROM 84, a gait issued in real time during walking (change of gait, etc.) is satisfied. Thus, a gait is generated in the mixed gait instantaneous value generator.

The standard gait is created by trial and error of various gaits off-line, and a gait having an end state in which the upper body does not diverge and continuous gait is guaranteed thereafter is selected from the gaits. , Stored in the ROM 84. For example, a series of continuous gaits, such as a gait that starts walking from a stationary state, an accelerated gait, a constant speed gait, a decelerated gait, and a stopped gait, are standard gaits in which continuous gait is guaranteed. This is a typical example.

A gait is described by gait parameters. The gait parameter is composed of a motion parameter and a ZMP parameter (more generally, a floor reaction force parameter). In this specification, the "floor reaction force parameter"
Is used to mean "parameters relating to the temporal pattern of the floor reaction force".

As shown in FIG. 7 and the like, the ZMP parameters are indicated by the position of the break point of the ZMP trajectory in the line graph and the passing time for the X, Y, and Z coordinates (directions) (only the X coordinate is shown). In the mixed gait instantaneous value generator, ZMP
The ZMP generator calculates the instantaneous value of the ZMP (trajectory) (the value of the current (current time) control cycle) based on the parameters.

The motion parameters include a foot (orbit) parameter and a body (orbit) parameter.

The foot trajectory parameters include an initial (at the time of leaving the floor) free leg position and posture, an end (at the time of landing) free leg position and posture, double leg support period time, single leg support period time, and the like. Based on these, the present applicant has previously disclosed in Japanese Patent Application Laid-Open Nos. 5-318339 and 5-324115
Using the technology proposed in Japanese Patent Publication No.
f, yf, zf) and attitude (θxf, θyf, θzf)
) Is obtained. Here, “posture” means “inclination or direction in space”.

The body trajectory parameters include a parameter for determining the posture of the body (the direction or inclination of the body 24 in space), a parameter for determining the body height (value in the Z direction), and an initial body position ( Displacement) and velocity parameters.

In the body horizontal position generator, a time function based on parameters for determining the position and posture of the body or a body function using the technique previously proposed by the present applicant in Japanese Patent Application Laid-Open No. 5-324115. The instantaneous values of the horizontal position (xb, yb) and attitude (θxb, θyb, θzb) are obtained. Note that, as described above, the position of the body means a representative point such as the position of the center of gravity of the body 24. In the body height determiner, the body height zb is determined by an appropriate method, for example, a method proposed by the present applicant simultaneously with this application.

Since the joints of the legged mobile robot 1 are composed of 12 joints as shown in FIG. 1, from the obtained positions and postures of both feet and the body position and posture, inverse kinematics calculation is performed. The target joint displacement is uniquely determined.
That is, the target posture of the robot this time is uniquely determined. (Accurately, the target joint displacement is calculated after the target positions and postures of both feet are corrected by the two-leg compliance control.)

Incidentally, originally, the gait is uniquely expressed only by the gait parameters. However, in this embodiment, in order to reduce the amount of calculation during walking, the body trajectory of the standard gait is preliminarily taken off-line. Is calculated and stored in a time series table, and the current time data is discharged in each control cycle. Although some errors occur, the ROM 84
When the capacity of the gait is small, the body trajectory of the standard gait may be approximated by a polynomial, and the coefficient of the approximate expression may be stored as a kind of gait parameter instead of the time series table.

Here, to supplement the description of the standard gait, the standard gait is more specifically expressed as a time function in this specification, and one gait is defined as g (t) as follows. When
It is represented by a set of f (t). g (t): Function describing motion (this function value represents the position / posture of the entire robot at time t. Specifically, the overall position / posture is, as shown in FIG. posture,
It is expressed by the position and posture of both feet). f (t): a function describing the ZMP trajectory (the value of this function represents the position (x, y, z) of the ZMP at time t, as described above).

The gait function is represented by a symbol, taking parameters into consideration. That is, both g (t) and f (t) are functions having gait parameters as parameters. In order to clearly indicate different gaits, all parameters should be explicitly described, but since there are so many types of parameters, the following notation that specifies only the parameter variables necessary for the following explanation is used It shall be. g (t: a, b, c, d, x (k), y (k)): Function describing motion trajectory f (t: a, b, c, d): Function describing ZMP trajectory (floor A function describing the reaction force) where: a: initial swing leg foot orientation (robot turning angle) b: initial swing leg foot front-back position (stride length) c: X coordinate of ZMP during single leg support period d: single leg support Y-coordinate of the ZMP of the period x (k): X-coordinate time series of body position y (k): Y-coordinate time series of body position

In this notation, time series is also treated as a kind of parameter. Originally, if the initial body position / velocity parameters are given, the gait is uniquely determined without a time series table of the body position. That is, the time series of the body position is redundant, but as described above, the time series table of the body position is also treated like a parameter describing the gait to reduce the calculation time of real-time gait generation. ing.

In this embodiment, the initial body position / velocity parameters also differ depending on the gait, but when the time series of the body position is specified, the first column of the time series is the initial body position / speed. -The description is omitted because it matches the speed parameter value.

In this embodiment, as described above, a plurality of types of gaits such as a gait that starts walking from a stationary state are prepared in advance as a standard gait.
Various gaits will be exemplified as specific examples. 0th standard gait g (t: a0, b0, c0, d0, x0 (k), y0 (k)) f (t: a0, b0, c0, d0) 1st standard gait g (t: a1, b0, c1, d1, x1 (k), y1 (k)) f (t: a1, b0, c1, d1) Second standard gait g (t: a0, b2, c2, d2, x2 (k), y2 (k)) f (t: a0, b2, c2, d2)

FIGS. 5 to 11 show the initial free leg foot position and posture and the last free leg foot position and posture of the standard gait. In order to express the gait, as shown in the figure, the origin of the coordinate system is a reference point where the support leg foot is in contact with the ground, the X-axis is the front-rear direction of the support leg foot, and the Y-axis is the left-right direction. .

Before starting to explain the flow chart of FIG.
Here, it is assumed that a request regarding the gait as shown in FIGS. 12 and 13 is given, and the required gait (the above-described approximate gait) from the above-described standard gait to satisfy the gait condition.
The operation of this device will be described by taking as an example the operation of generating a (third gait). The third gait has an arbitrary initial free leg foot orientation a3 and an arbitrary initial free leg foot front-back position b3.

Of course, the value of the parameter a of the third gait is a
The value of 3, b is b3. The problem is that the other parameters c and d and the table x
(k) and y (k) are determined.

Then, the third gait is set as g (t: a3, b3, c3, d3, x3 (k), y3 (k)) f (t: a3, b3, c3, d3), and the parameter c3, d3 and time series table x3 (k),
Determine y3 (k).

Here, the terms are defined as follows:
And b are the required setting values for this device from outside (by operating the joystick 62 or the like by the operator) other than this device or another device (such as the two-leg compliance controller in the block diagram of FIG. 3). Request) contains the value entered as the same. Such a parameter that requires a certain set value directly is called a direct setting parameter.

In this embodiment, for simplification of the description,
Only the initial swing leg foot direction and the front / rear position (step length) can be arbitrarily changed, but other than those input as required setting values are as described above. 1) Instructions from the operator (turning angle and angular velocity) Command, moving speed command, etc.) 2) If another processing device (such as a track guidance device) is provided,
3) Boundary conditions with the previous and / or next gait (initial body position and posture, initial free leg foot position and posture, etc.) Can be mentioned. 3) will be described in the fourth embodiment.

On the other hand, parameters such as parameters c and d and tables x (k) and y (k) whose values are determined so as to satisfy gait conditions in accordance with the values of directly set parameters (or table ) Are referred to as dependent parameters (or dependent tables). Since it is extremely difficult to directly set the time series table, the time series table hardly serves as a direct setting parameter.

The first and second gaits other than the zeroth standard gait
The standard gait is obtained by changing only one of the directly set parameters (here, parameters a and b) based on the 0th standard gait. Thus, the gait (the 0th standard gait) serving as the core of the selected standard gait is called a base gait.

Here, the method of obtaining the parameters of the required gait (the parameters of the required gait is also referred to as “mixed parameters” in this specification) will be described. The dependent parameters c3 and d3 and the dependent time series table x3 (k) , y3 (k) directly change according to the setting parameters a and b in order to satisfy the condition of the gait. Moreover, it is considered that these values change continuously with changes in the parameters a and b. That is, they are continuous functions of parameters a and b. Therefore, the following approximation principle can be used. That is, the parameter of the required gait can be obtained by performing the approximate calculation.

If the function P is a smooth continuous function, P (a0 + Δa, b0 + Δb) = P (a0, b0) + δP / δa | (a = a0, b = b0) * Δa + δP / δb | (a = a0, b = b0) * Δb ・ ・ ・ Approximation formula 1 However, δP / δa | (a = a0, b = b0): bias of P at a = a0, b = b0 Derivative value δP / δb | (a = a0, b = b0): Partial differential value of P with respect to b at a = a0, b = b0

Further, δP / δa | (a = a0, b = b0) = (P (a0 + Δa1, b0) -P (a
0, b0)) / Δa1 δP / δb ｜ (a = a0, b = b0) = (P (a0, b0 + Δb2)-P (a
0, b0)) / Δb2, the approximate expression 1 is P (a0 + Δa, b0 + Δb) = P (a0, b0) + (P (a0 + Δa1, b0) -P (a0, b0)) / Δa1 * Δa + (P (a0, b0 + Δb2) -P (a0, b0)) / Δb2 * Δb ・ ・ ・ Approximation formula 2

The first embodiment is obtained by developing the approximation formula 2, and the second embodiment described later is obtained by developing the approximation formula 1.

Continuing the description of the method of obtaining the parameters of the required gait, specifically, by internally or externally dividing all dependent parameters and all dependent time series tables as shown in the following equation, The parameters of the third gait are obtained from the standard gait. This is because the direct setting parameter and the dependent parameter can be considered to be in a proportional relationship.

If a3 is inside a1 and a2, it is an internal part, and if a3 is outside it, it is an external part. In addition, some parameters and some time series tables in which the difference can be ignored dynamically may not be divided internally or externally, and may be the base gait. Further, regarding the time series table, if all data at all times are calculated at one time, the load on the first arithmetic unit 80 is too large.

The following shows a mixed expression (depending on the internal or external division) of the dependent parameter and the dependent table (this mixed expression is hereinafter referred to as “expression 1”). c3 = c0 + (c1-c0) * (a3-a0) / (a1-a0)
+ (c2-c0) * (b3-b0) / (b2-b0) d3 = d0 + (d1-d0) * (a3-a0) / (a1-a0)
+ (d2-d0) * (b3-b0) / (b2-b0) x3 (k) = x0 (k) + (x1 (k)-x0 (k)) * (a3-a0)
/ (a1-a0) + (x2 (k)-x0 (k)) * (b3-b0) /
(b2-b0) y3 (k) = y0 (k) + (y1 (k)-y0 (k)) * (a3-a0)
/ (a1-a0) + (y2 (k)-y0 (k)) * (b3-b0) /
(b2-b0)

The above equation 1 is a weighted average (weighted average) of each standard parameter and table value. For example, if the formula for c3 is transformed, c3 = (1-(a3-a0) / (a1-a0)-(b3-b0
) / (b2-b0)) * c0 + (a3-a0) / (a1-a0
) * C1 + (b3-b0) / (b2-b0) * c2, and the sum of the coefficients of c0, c1, and c2 is 1.
c3 has weights for c0, c1 and c2, respectively (1-(a3-a0) / (a1-a0)-(b3-b0) /
(b2-b0)) (a3-a0) / (a1-a0) (b3-b0) / (b2-b0) weighted average.

When it is desired to set the terminal body position / velocity to an arbitrary value as a directly set parameter, it is only necessary to select some standard gaits with different ZMP parameters from the base gait and perform weighted averaging. Since the terminal state of the mixed gait substantially matches the weighted average of the terminal state of each standard gait to be mixed, all of the terminal states of the mixed gait (the position and speed of the upper, lower, left, and right sides of the body) are required values. Thus, the weight of the weighted average may be obtained.

When the body position is obtained from the time series table, the termination conditions completely match. However, the seventh
As described in the embodiment, when the body position is sequentially obtained so as to satisfy the ZMP set trajectory by the dynamics calculation, the position is slightly shifted.

At this time, in order to determine the weight,
Since the end state of each standard gait to be mixed is required, the processing time can be reduced by using a table in which the end state of each standard gait is stored in advance.

Determining the parameters of the mixed gait by the above method, and then generating or generating the instantaneous value of the required gait at each time by the same gait generation algorithm as the standard gait created off-line Can be.

The operation of this apparatus will be described below with reference to the flow chart of FIG. 4. Assume that the data group of the standard gait created off-line in S10 is stored in the ROM 84 in advance. Next, the process proceeds to S12, in which the end body position and the speed of each standard gait are obtained and stored.

As described above, the position of the body 24 is determined by the function x
Since it is tabulated as (k) and y (k), the speed is obtained from the data at the end of the table and the time immediately before it. Specifically, the value at the last time is set as the position, and the difference between the value at the last time and the value at the time immediately before is set as the speed.

Next, the process proceeds to S14, in which the timer value t is set to 0 (timer is started), the process proceeds to S18 via S16, and waits for a timer interrupt (different from the timer of S14). The flow chart of FIG. 4 is activated, for example, every 20 ms. That is, the control cycle is every 20 ms.

Then, the process proceeds to S20, where it is determined whether or not the gait is a gait switch, specifically, a one-step switch. If affirmative, the process proceeds to S22, where the timer is reset to 0, and the process proceeds to S24. Then, the required value for the gait (specifically, the required values (a and b described above) of the direct setting parameters of the current time's gait) is read.

Next, the process proceeds to S26, where the gait is mixed.

FIG. 14 is a subroutine flowchart showing the operation.

First, in S100, a standard gait in which the value of the directly set parameter is close to the required value is selected as a base gait. Referring to FIG. 15, the standard gait in which the values of the directly set parameters a and b are close to the required values (third gait) a3 and b3 is the zeroth standard gait. Select as

Next, proceeding to S102, focusing on the directly set parameters, one standard gait in which only one of the parameters differs from the base gait is selected one by one. If there are a plurality of candidates, the parameter value of the candidate should be as close as possible to the value of the parameter of interest, and if possible, the one where the required value is located between the parameter value of the base gait and the parameter value of the candidate select. For example, in the case as shown in FIG. 16, the standard gait candidate 1 is adopted for the parameter a.

As will be described later in the fifth embodiment, when the mass of the leg link 2 is sufficiently smaller than that of the body 24, the end position of the free leg foot position in the directly set parameters.・ Even if the posture parameters are slightly changed, the robot 1
The inertia force generated is almost the same. In other words, it can be considered that even if the parameter value is changed, other parameters are hardly affected in satisfying the dynamic equilibrium condition.

Therefore, if the end free leg foot position / posture parameters are slightly different from the base gait, the parameter values of the base gait may be rewritten as required values.
Therefore, it is not necessary to select a gait that differs from the base gait only for the end free leg foot position / posture parameters for mixing.

Then, the program proceeds to S104, in which a required value is substituted for the directly set parameter of the mixed gait, and to S106, the dependent parameter value of the mixed gait is calculated by the above-described parameter calculation formula (Equation 1). Substitute for the parameter.

Returning to FIG. 4, the program proceeds to S28, in which the mixed gait parameter (table) is substituted for the desired gait parameter (table), and then to S30, the instantaneous value of the desired gait (the value of the current control cycle) ) Is calculated.

FIG. 17 is a subroutine flowchart showing the operation.

First, in step S200, the ZMP at time t is determined based on the mixed gait parameters, and the flow advances to step S202 to determine the position and orientation of both feet at time t using the proposed technique based on the mixed gait parameters.

Then, the program proceeds to S204, in which the body height at time t is determined from both foot positions and postures at time t and the upper body horizontal position at time t-Δt. The details are described in another application proposed at the same time as described above, and do not have a direct relation to the gist of the present invention, so that the description thereof will be omitted.

Subsequently, the flow proceeds to S206, where the mixed gait at time t is calculated according to the above-described mixing formula (Equation 1) based on the body position of the selected standard gait (including the base gait) at time t. Calculate the horizontal position.

Subsequently, the flow returns to S32 of the flow chart of FIG. 4 to update the time t by Δt, and returns to S18 to repeat the above processing.

The two-leg compliance controller performs the two-leg compliance based on the gait generated as described above, then calculates a target joint angle, and controls the drive toward the target value.

The two-leg compliance control will be briefly described with reference to the flow chart of FIG.
At 300, the two-leg compliance operation amount is calculated.

FIG. 19 is a subroutine flowchart showing this operation. In step S400, the detection value of the six-axis force sensor 44 is read, and the flow advances to step S402 to measure the actual ZMP position (actual floor reaction force center point) from the detection value. And S404
, The deviation, that is, the deviation direction and the deviation amount X are obtained by comparing with the ZMP target position, the flow proceeds to S406, and the operation amount of the foot is calculated according to the illustrated formula. The flow proceeds to S408, and the target position / posture of both feet is obtained. Is shifted according to the obtained operation amount. Since the actual ZMP position and the actual floor reaction force center point coincide, they are treated as synonymous in this specification.

Returning to the flow chart of FIG. 18, the program then proceeds to S302, in which a known inverse kinematics calculation is performed based on the foot position / posture (corrected values when corrected by compliance control) and the body position / posture. Through 12
After calculating the target angles of the individual joints, and proceeding to S304, when it is determined from the output of the tilt sensor 11 that the robot 1 is unstable, the target posture is corrected so that the floor reaction force center point is intentionally shifted. Next, the routine proceeds to S306, in which all the joints are controlled to follow the target angle. This is specifically performed by the second arithmetic unit 82.

Since this embodiment is configured as described above, the gait can be freely and in real time generated to realize an arbitrary stride and turning angle. Further, the legged mobile robot can be freely walked while being driven and controlled to the target joint angle based on the gait generated in real time.

More specifically, a base gait is appropriately selected from several standard gaits stored for an arbitrary request for a stride and a turning angle, and a mixing operation such as synthesis or deformation, ie, a mixing operation, is performed. By performing the above-described approximation calculation, it is possible to generate an approximate gait satisfying the above-mentioned gait conditions and gait requirements in real time.

Further, with the above-described configuration, it is possible to set a delicate angle and stride length, and to mix the existing standard gaits to some extent (if the boundary state is close) to mix them with each other. It is also possible to arbitrarily generate a realistic gait.

Further, the configuration described above can reduce the amount of calculation. Specifically, the amount of calculation of the first calculation device 80 is reduced to about 1/10 compared to solving the dynamics in real time. Furthermore, since the optimal standard gaits are stored and mixed to generate gaits,
The capacity of the memory (ROM 84) can also be reduced. Further, since the optimality is satisfied, a gait satisfying the conditions 1) to 5) can be easily obtained without trial and error.

FIG. 20 is a flow chart showing a second embodiment of the present invention.

In the second embodiment, the sensitivity for the gait parameters is used together with the standard gait. In general, “sensitivity” means the degree to which the output or characteristics of the system are affected by changes in inputs, parameters, or environmental conditions, but in the second embodiment, the unit change amount of a certain gait parameter The amount of change of other parameters per hit (hereinafter referred to as “gait parameter sensitivity”) is obtained, and the amount of change (or amount of change) of the parameter is obtained using the obtained amount.

More specifically, the gait parameter sensitivity is
In a standard gait, directly set parameters (for example, first
When the initial free leg direction “a” and the initial free leg front-back position “b” of the embodiment are changed by a small amount, in order for the gait to become a desirable standard gait, how many other dependent parameters are required Used to indicate that it must be changed.
Specifically, the gait parameter sensitivity is defined as a change amount of the dependent parameter per unit change of the directly set parameter.

As a result, in the second embodiment, the number of standard gaits is reduced, and only those that become a base gait are used. That is, for example, a series of continuous base gaits such as a gait to start walking from a stationary state, an acceleration gait, a constant speed gait, a deceleration gait, and a stop gait are prepared. A staggered standard gait was not prepared.

More specifically, the sensitivity of the parameter q to the parameter p in a certain standard gait i is described as Di (q, p). Similarly, in the time series table, the sensitivity of the k-th data x (k) to the parameter p is described as Di (x (k), p).

A method of generating a required gait according to the second embodiment will be described.

The relationship between the sensitivity D0 (c, a) of the parameter c with respect to the parameter a in the 0th standard gait and the first standard gait is as follows. If (a1-a0) is small enough, by definition, D0 (c, a) = (c1-c0) / (a1-a0) D0 (d, a) = (d1-d0) / (a1-a0) D0 (x (k), a) = (x1 (k)-x0 (k)) / (a1-a0) D0 (y (k), a) = (y1 (k)-y0 (k)) / ( a1-a0).

Similarly, the relationship between the sensitivity D0 (c, b) of the parameter c with respect to the parameter b in the 0th standard gait and the second standard gait is as follows. (b2-b0)
Is sufficiently small, by definition, D0 (c, b) = (c2-c0) / (b2-b0) D0 (d, b) = (d2-d0) / (b2-b0) D0 (x (k ), b) = (x2 (k)-x0 (k)) / (b2-b0) D0 (y (k), b) = (y2 (k)-y0 (k)) / (b2-b0) is there.

By substituting these into the mixed equations of the parameters and tables in the first embodiment, the following equation (hereinafter referred to as “equation 2”) is obtained as a parameter determination equation (using parameter sensitivity).

C3 = c0 + D0 (c, a) * (a3-a0)
+ D0 (c, b) * (b3-b0) d3 = d0 + D0 (d, a) * (a3-a0) + D0 (d, b
) * (B3-b0) x3 (k) = x0 (k) + D0 (x (k), a) * (a3-a0) + D
0 (x (k), b) * (b3-b0) y3 (k) = y0 (k) + D0 (y (k), a) * (a3-a0) + D
0 (y (k), b) * (b3-b0)

Referring to the flow chart of FIG. 20 on the premise of the above description, the base gait and gait parameter sensitivity created off-line in S500 are stored, and S502 to S514 correspond to the first embodiment. The same processing is performed, and the process proceeds to S516, where gait mixing processing is performed.

FIG. 21 is a subroutine flow chart showing the operation. In step S600, a base gait whose directly set parameter value is close to the required value is selected, and the flow advances to step S602 to request the directly set parameter of the mixed gait. Assign a value,
Proceeding to S604, dependent parameters are determined according to Equation 2. Note that the time series table of the body horizontal position may be calculated at one time, but since the amount of calculation is large, only a necessary value may be calculated for each control cycle in an instantaneous value generation subroutine described later.

Returning to FIG. 20, the process proceeds to S520 via S518 and S510, and the instantaneous value of the desired gait is calculated.

FIG. 22 is a subroutine flow chart showing the operation, and is similar to the first embodiment.
The processing from 00 to S706 is performed to calculate the position and orientation of the foot and the body. In step S706, the body horizontal position is calculated in accordance with the expression relating to x and y in Expression 2. Subsequently, returning to FIG. 20, the processing returns to S508 via S522, and the above processing is repeated.

Since the second embodiment is configured as described above, the same effects as those of the first embodiment can be obtained.

However, the superiority of the second embodiment over the first embodiment differs depending on the combination of the standard gaits to be stored and the properties of the standard gaits. For example, when the entire space created by a set of direct setting parameters is referred to as a parameter space, in the first embodiment, each of the direct setting parameters is discretized at a certain interval, and a grid-like discontinuous portion is formed in the parameter space. Create a space and create the desired standard gait for all directly set parameter sets of the discontinuous subspace. That is, the values of the parameters (dependent parameters) other than the directly set parameters are adjusted so as to obtain a desired gait.

If the standard gait created in this manner is provided, most of the standard gaits satisfy the condition as the base gait (the condition of the base gait is changed from the gait to all the parameters to be changed). On the other hand, there is a standard gait that differs only in one of them). Therefore, an arbitrary gait can be created by the method of the first embodiment, and there is no need to have parameter sensitivity.

In the second embodiment, if the parameter sensitivities are substantially equal for a wide range of gaits, 1
Instead of having parameter sensitivity for each gait, the gait may be shared in a wide range of gaits. By doing so, the capacity of the memory (ROM 84) can be saved, and in that case, a much smaller memory capacity is sufficient as compared with the first embodiment.

For example, D0 (c, a), D1 (c, a), D2 (c,
a),... Dn (c, a) are almost the same value, and the average value is D (c, a), and D0 (c, a) in the parameter determination formula using the parameter sensitivity described above. , D1 (c, a), .. may be used in common with D (c, a).

FIG. 23 is a flow chart showing a third embodiment of the present invention.

In the third embodiment, the standard gait similar to that of the first embodiment is used, and the gait parameter sensitivity of the second embodiment is also used.

A method of generating a required gait according to the third embodiment will be described.

In the third embodiment, in determining the value of the dependent parameter, the influence of the direct setting parameter is partially used in the method of the first embodiment, and the rest is used in the method of the second embodiment. Is determined using That is, mixing of gaits and addition of perturbations due to gait parameter sensitivity are used together.

For example, using the method of the first embodiment for the effect of a in the direct setting parameters and the method of the second embodiment for the effect of b, the following equation is obtained. . c3 = c0 + (c1-c0) * (a3-a0) / (a1-a0)
+ D0 (c, b) * (b3-b0) d3 = d0 + (d1-d0) * (a3-a0) / (a1-a0)
+ D0 (d, b) * (b3-b0) x3 (k) = x0 (k) + (x1 (k)-x0 (k)) * (a3-a0)
/ (a1-a0) + D0 (x (k), b) * (b3-b0) y3 (k) = y0 (k) + (y1 (k)-y0 (k)) * (a3-a0)
/ (a1-a0) + D0 (y (k), b) * (b3-b0)

Referring to FIG. 23, the standard gait and gait parameter sensitivity created off-line in S800 are stored, and the process proceeds to S802 and thereafter to perform the same processing as in the first embodiment. Proceeding to the flow chart of FIG. 24, gaits are mixed in and after S900.
Further, the process proceeds from S820 in the flow chart of FIG. 23 to S1000 and later in the flow chart of FIG. 25 to calculate an instantaneous value.

Since the third embodiment is configured as described above, it has the same effect as the previous embodiment.

FIG. 26 is a flowchart showing a fourth embodiment of the present invention.

As shown in FIG. 27, the mixed gaits (required gaits) obtained from the standard gait tend to slightly deviate in the body position and speed at the gait boundary. The body position / velocity is the initial body position /
It is necessary to adjust to the speed. Therefore, in the fourth embodiment, the boundary condition of the body position / velocity is arbitrarily set. That is, the initial body position / velocity and the end body position / velocity are also considered directly as setting parameters,
The speed was made continuous in boundary conditions.

In summary, the direct setting parameters except for the initial body position / velocity and the end body position / velocity are all set as desired using the methods of the first to third embodiments. Then, the ZMP parameters are corrected so that the initial body position / velocity and the end body position / velocity are also as desired.

The processing in the first half is the same as that in the previous embodiment, so that the description is omitted. The ZM is set so that the initial body position / velocity and the end body position / velocity in the latter half are also as desired.
A method for correcting the P parameter will be described.

In the gait generation method described above, it is assumed that gait parameters other than the initial body position / velocity, end body position / velocity, and ZMP parameters are fixed.

In this case, when the initial body position / velocity and the ZMP pattern are set, the gait satisfying the dynamic equilibrium condition is uniquely determined, so that the terminal body position / velocity is also uniquely determined. . Conversely, when the terminal body position / velocity and the ZMP pattern are given, the initial body position / velocity is uniquely determined. Therefore, in order to arbitrarily set the initial body position / velocity and the end body position / velocity, the ZMP pattern must be adjusted accordingly. That is, once the ZMP pattern is determined, the initial body position / velocity and the end body position / velocity have a one-to-one mapping relationship.

In the following, for the sake of simplicity, let us consider only the behavior of the upper body in the front-rear direction (X direction).
Since the state quantity is two variables of position and velocity, in order to be able to arbitrarily set the boundary conditions relating to these, Z
The MP pattern requires two or more parameters.

Here, the ZMP pattern and its parameters are shown in FIG. 2 so that the body boundary position / velocity can be largely changed by only slightly changing a part of the ZMP parameters.
Set as 8

This corresponds to the ZMP used in the previous embodiment.
The pattern is more complicated than the pattern. Because, using the same ZMP pattern as used in the previous embodiment, in order to simultaneously satisfy both the position and the velocity of the upper body in the front-rear direction, the foot trajectory is affected in addition to the parameter c. This is because the parameter t2 for the non-existing time must be changed, but even if t2 is largely changed, the body boundary position / velocity can be corrected only slightly.

In the fourth embodiment, the ZMP parameter can be determined by using any of the methods of the first embodiment and the second embodiment, but here, the second embodiment is used. The correction amount of the ZMP parameter and the correction trajectory of the body are determined by the following procedure using the above method. Note that the correction amount of the ZMP parameter means a correction amount added to the ZMP parameter of the mixed gait to set the upper body boundary condition of the mixed gait to a desired value.

The operation of the fourth embodiment will be described with reference to FIG. 26. In S1100, a standard gait created off-line is stored in the same manner as in the previous embodiment.
Proceeding to S1102, the end body position and speed of various standard gaits are obtained and stored, and then the procedure proceeds to S1104 to perform the following processing.

First, paying attention to the ZMP parameter c in the mixed gait, this is slightly perturbed by Δc as shown in FIG. 29, and the time series data of the body position is thereby adjusted so as to satisfy the dynamic equilibrium condition.摂 x (k) is obtained.

At this time, Rmix (x (k), c) = Δx (k) / Δc. This Rmix (x (k), c) is called the partial differential sensitivity of x (k) to c (corresponding to the gait parameter sensitivity). Similarly, as shown in FIG. 30, Rmix (x (k), e) is obtained for the ZMP parameter e.

In the second embodiment, when obtaining the sensitivity of x (k) to a certain noted parameter, the noted parameter and x (k) are determined while maintaining the optimum gait as a constraint. Excluding several dependent parameters were adjusted.
On the other hand, in the fourth embodiment, the sensitivity is obtained by ignoring the optimality of the gait and fixing the parameter of interest and all parameters except x (k). In other words, in the second embodiment, the total derivative is obtained by using the optimality as a constraint, and in the fourth embodiment, the partial differential is obtained by ignoring the optimality.

Next, the flow advances to S1106, where the terminal body position of the terminal body with respect to the ZMP parameter c in the mixed gait is determined from the terminal points of Rmix (x (k), c) and Rmix (x (k), e) and the values immediately before them. Partial differential sensitivity Rmix (Xe, c), partial differential sensitivity R of terminal body velocity
Mix (Ve, c), partial differential sensitivity Rmix (Xe, e) of the terminal body position with respect to the ZMP parameter e, and partial differential sensitivity Rmix (Ve, e) of the terminal body velocity are calculated by the following equations.

Rmix (Xe, c) = Rmix (x (ke), c) Rmix (Ve, c) = (Rmix (x (ke), c) -Rmix (x (ke-1), c))
/ Δt Rmix (Xe, e) = Rmix (x (ke), e) Rmix (Ve, e) = (Rmix (x (ke), e)-Rmix (x (ke-1),
e)) / Δt where ke is the last array number of the time series, and Δt is the step time of the time series (discrete sampling time).

Then, the flow advances to S1108, where the element in the first row and the first column is Rmix (Xe, c), the element in the first row and the second column is Rmix (Xe, e), and the element in the second row and the first column is obtained from the obtained sensitivity. Element is Rmix (Ve,
c) Create a 2 * 2 matrix in which the elements in the second row and second column are Rmix (Ve, e).

Then, the process proceeds to S1110, where an inverse matrix of the created matrix is obtained. The element in the first row and first column of this inverse matrix is
The partial differential sensitivity Rmix (c, Xe) of the parameter c to the terminal body position Xe, and the element in the second row and first column is the terminal body position X
partial differential sensitivity Rmix (e, Xe) of parameter e to e,
The element in the first row and the second column is a partial differential sensitivity Rmix (c, Ve) of the parameter c with respect to the terminal body velocity Ve, and the element in the second row and the second column is a partial deviation of the parameter e with respect to the terminal body velocity Ve. Let the differential sensitivity be Rmix (e, Ve).

If the sensitivity hardly changes in any mixed gait, the average value is calculated as R (c, Xe), R (e, Xe), R (c, Ve), R ( e, Ve)
It may be used commonly for all gaits.

Then, the flow advances to S1112, where attention is paid to the initial body position in the mixed gait, which is slightly perturbed by Δx0 as shown in FIG. 31, and accordingly, the body position is adjusted so as to satisfy the dynamic equilibrium condition. Perturbation Δ of the time series data of
Find x (k). Naturally, Δx (0) = Δx0.

At this time, Rmix (x (k), x0) = Δx (k) / Δx0. Similarly, as shown in FIG. 32, only the initial body velocity is slightly perturbed by Δv0 to obtain Rmix (x (k), v0).

Next, proceeding to S1114, the calculated sensitivity Rmix
Based on the end of (x (k), x0) and the end of Rmix (x (k), v0) and the value immediately before the end, the partial differential sensitivity Rmix (Xe , x0), partial differential sensitivity of terminal body velocity Rmix (Ve, x0), partial differential sensitivity of terminal body position relative to initial body velocity Rmix (Xe, v0), partial differential sensitivity of terminal body velocity Rmix (Ve, v0) is obtained by the following equation.

Rmix (Xe, x0) = Rmix (x (ke), x0) Rmix (Ve, x0) = (Rmix (x (ke), x0)-Rmix (x (ke-1),
x0)) / Δt Rmix (Xe, v0) = Rmix (x (ke), v0) Rmix (Ve, v0) = (Rmix (x (ke), v0)-Rmix (x (ke-1),
v0)) / Δt where, ke: last sequence number of the time series, Δt: time step of the time series.

Subsequently, the flow advances to S1116 to reset and start the timer to 0, and then to S1120 via S1118.
To wait for a timer interrupt, proceed to S1122, determine whether or not the gait is at a switch, and if affirmative, proceed to S1124 to reset the timer to 0, and then proceed to S1126 to directly set the parameter Is read, and the process advances to S1128 to mix gaits.

That is, the process proceeds to S1200 of the subroutine flow chart of FIG. 33 to create a mixed gait parameter using the same method as in the first to third embodiments.
At this time, an initial body position / velocity and an end body position / velocity that are close to the required values are selected as base gaits.

However, after selecting the base gait, the initial body position / velocity and the terminal body position / velocity are excluded from the directly set parameters, and a mixed gait is generated in the same manner as in the previous embodiment. .

Then, the process advances to S1202, where the initial body position
The difference between the required speed value and the initial body position / velocity of the mixed gait is determined, and these are defined as Δx0 and Δv0. Similarly, the difference between the required end body position / velocity value and the end body position / velocity of the mixed gait is determined, and these are defined as ΔXe and ΔVe.

Next, the flow advances to S1204, where only the initial body position / velocity is changed to the required initial body position / velocity value in the mixed gait, and accordingly, the body position is changed so as to satisfy the dynamic equilibrium condition. The perturbation amounts ΔXse and ΔVse of the terminal body position / velocity of the gait in which the time-series data is perturbed are obtained by the following equations. ΔXse = Rmix (Xe, x0) * Δx0 + Rmix (Xe, v0) * Δv0 ΔVse = Rmix (Ve, x0) * Δx0 + Rmix (Ve, v0) * Δ
v0

In the gait of S1204, the initial body position / velocity coincides with the required value, but the difference between the required end body position / velocity value and the terminal body position / velocity of this gait is ΔXe + Δ
Xse, ΔVe + ΔVse. Then, the process proceeds to S1206, and the correction amounts Δc and Δe of the ZMP parameter for setting the difference to 0 are obtained by the following equations. Δc = Rmix (c, Xe) * (ΔXe + ΔXse) + Rmix (c,
Ve) * (ΔVe + ΔVse) Δe = Rmix (e, Xe) * (ΔXe + ΔXse) + Rmix (e,
Ve) * (ΔVe + ΔVse)

Then, the flow advances to S1208 to determine ZMP parameters c and e by the following equations. c = cmix + Δce = emix + Δe where cmix: the value of the parameter c of the mixed gait, and emix: the value of the parameter e of the mixed gait.

Next, returning to the flow chart of FIG.
Proceeding to 1130, the parameters of the mixed gait are substituted for the parameters of the desired gait. However, the ZMP parameter c,
The value obtained in S1208 is substituted for e. Then S1
Proceeding to S1132 via 122, the instantaneous value of the desired gait is calculated.

Specifically, the flow advances to S1300 in the subroutine flow chart of FIG. 34 to obtain the ZMP at time t by the same method as in the previous embodiment, and the flow advances to S1302 to execute the same method as in the previous embodiment. Then, the positions and postures of both feet at time t are obtained, and the flow advances to S1304 to obtain the body position of the desired gait by the following equation. x (k) = xmix (x) + Rmix (x (k), c) * Δc + Rmix (x
(k), e) * Δe + Rmix (x (k), x0) * Δx0 + Rmix (x (k), v0)
* Δv0 where xmix (x) is the k-th data of the body trajectory time series of the mixed gait. Although the description is omitted, it goes without saying that the body position in the Y direction is calculated in the same manner. The body position in the Z direction may be calculated in the same manner as the values in the X and Y directions, or may be calculated by a method proposed at the same time as this application.

By the above processing, the waveforms shown in FIG. 27 and FIGS. 29 to 32 are superimposed, thereby generating a desired gait having desired boundary conditions as shown in FIG. be able to. FIG.
In the situation shown in FIG. 5, c <0, c + Δc <0, e <0, e + Δe> 0.

Since the fourth embodiment is configured as described above, it is possible to generate gaits freely and in real time, and to make the displacement and the displacement speed of each part of the robot continuous at boundaries between the generated gaits. Can be.

FIG. 36 is a flowchart showing a fifth embodiment of the present invention.

In the fourth embodiment, the partial differential sensitivity Rmix (x (k), c) of the body position time series with respect to the ZMP parameter c and the body position time series with respect to the ZMP parameter e in the mixed gait are described. , The partial differential sensitivity Rmix (x (k), e) of the body position time series relative to the initial body position Rmix (x
(k), x0), and the partial differential sensitivity Rmix (x (k), v0) of the body position time series with respect to the initial body velocity. So to speak, Z
It can be said that the model of the perturbation of the upper body trajectory with respect to the perturbation of MP was in the form of a time series table.

In the fifth embodiment, a robot linear perturbation dynamics model (Z
Using a linearized model representing the relationship between the perturbation of the body position velocity and the MP perturbation, the boundary conditions relating to the body position / velocity were made to match the required values.

Hereinafter, the basic principle of this method will be described.
In the following description, only the body behavior and the front-back direction (X direction) component of the ZMP trajectory will be discussed, but the left-right direction (Y
The direction) may be similarly considered.

It is assumed that the mass of the upper body 24 of the robot 1 is sufficiently larger than the mass of the other parts, and that the vertical acceleration of the upper body during walking is small. When this special condition is satisfied, the relation between the ZMP perturbation and the perturbation of the body position / velocity, that is, the perturbation dynamics model is approximated by the following equation and the linear inverted pendulum model shown in FIG. As shown, in this model, the fulcrum 22a is also movable.

Dxmdl / dt = vdvmdl / dt = ω0 * (xmdl-ZMPmdl) where xmdl is the position of the center of gravity of the linear inverted pendulum, and vmdl is the center of gravity of the linear inverted pendulum. Ω0 is a positive constant,
It has the following relationship with the upper center of gravity height h and the gravitational acceleration g. ω0 = √ (g / h)

When the above special conditions are satisfied, the highly perturbed dynamic models Rmix (x (k), c), Rmix (x (k), e), Rmix (x (k), x0), R
Instead of mix (x (k), v0), the above-described linear inverted pendulum model can be used. Hereinafter, when it is desired to clearly indicate the time t in the notation of the model state, the notation is expressed as xmdl (t) and vmdl (t).

Description will be made with reference to the flow chart of FIG. 36 on the premise of the above. The structure of the ZMP pattern is the same as that of the fourth embodiment.

First, in step S1400, the standard gait is stored, and the flow advances to step S1402 to calculate and store the terminal body position / velocity at the end of each standard gait.
In the P parameters, c and e are c = 1, e = 0, and p
(ZMP value at time 0 as shown in FIG. 35) =
0, the value of ZMP at other break points is 0, and the initial state is x0 =
0, v0 = 0, calculate the behavior of the perturbation dynamic model up to the end time of the gait, and calculate the end model position Xe and the end model speed Ve
Ask for.

Specifically, the behavior of the model is obtained by using the ZMP pattern correction amount based on Δc when Δc in FIG. 29 is set to 1 as a model input, and the model end state is obtained.
Subsequently, the partial differential sensitivity of the terminal model position Xe with respect to the ZMP parameter c is obtained by the following equation. R (Xe, c) = Xe

Similarly, the partial differential sensitivity of the terminal model speed Ve with respect to the ZMP parameter c is obtained by the following equation. R (Ve, c) = Ve

The behavior of the perturbation dynamics model may be calculated by discretizing the model and then sequentially calculating or solving it analytically.

Then, the flow advances to S1406, where c and e in the ZMP parameters are c = 0, e = 1, p = 0, the ZMP value at other break points is 0, and the initial state is x0 = 0, v0 = 0
Until the end time of the gait, the behavior of the perturbation dynamic model is calculated, and the end model position Xe and the end model speed Ve are obtained.

Next, the partial differential sensitivity of the terminal model position Xe with respect to the ZMP parameter e is obtained by the following equation. R (Xe, e) = Xe

Similarly, the partial differential sensitivity of the terminal model speed to the ZMP parameter e is obtained by the following equation. R (Ve, e) = Ve

Then, the flow advances to S1408, where c and e in the ZMP parameters are c = 0, e = 0, p = 0, the ZMP value at other break points is 0, and the initial state is x0 = 1, v0 = 0
Then, the behavior of the perturbation dynamic model is calculated until the end time of the gait, and the end model position Xe and the end model speed Ve are obtained.

Next, the partial differential sensitivity of the terminal model position Xe with respect to the initial model position x0 is obtained by the following equation. R (Xe, x0) = Xe

Similarly, the partial differential sensitivity of the terminal model speed Ve with respect to the initial model position x0 is obtained by the following equation. R (Ve, x0) = Ve

Then, the flow advances to S1410, where c and e in the ZMP parameters are c = 0, e = 0, p = 0, the ZMP value at other break points is 0, and the initial state is x0 = 0, v0 = 1
Until the end time of the gait, the behavior of the perturbation dynamic model is calculated, and the end model position Xe and the end model speed Ve are obtained.

Next, the partial differential sensitivity of the terminal model position Xe with respect to the initial model speed v0 is obtained by the following equation. R (Xe, v0) = Xe

Similarly, the partial differential sensitivity of the terminal model speed Ve with respect to the initial model speed v0 is obtained by the following equation. R (Ve, v0) = Ve

Then, the flow advances to S1411, where c = 0, e =
0, p = 1, ZMP at other break points is 0, initial state is
Assuming that x0 = 0 and v0 = 0, the behavior of the perturbation model up to the end time of the gait is calculated, and the partial differential sensitivity of the end model position Xe and the end model speed Ve is obtained.

Next, the partial differential sensitivity of the terminal model position Xe with respect to the initial model speed v0 is obtained by the following equation. R (Xe, p) = Xe

Next, the partial differential sensitivity of the terminal model speed Ve with respect to the initial model speed v0 is obtained by the following equation. R (Ve, p) = Ve

Then, the flow advances to S1412, where the element in the first row and first column is R (Xe, c), the element in the first row and second column is R (Xe, e), and the second row and first column are obtained from the obtained sensitivity. Is R (Ve, c) and the second element
Create a 2 * 2 matrix with R (Ve, e) elements in the row and second column.

Then, the flow advances to S1414, where an inverse matrix of the created matrix is obtained. The element in the first row and first column of the inverse matrix is the partial differential sensitivity R of the parameter c with respect to the terminal model position Xe.
(c, Xe), the element in the second row and the first column is the terminal model position Xe
Partial sensitivity R (e, Xe) of parameter e to
The element in the row and the second column is the partial differential sensitivity R (c, Ve) of the parameter c to the terminal model velocity Ve, and the element in the second row and the second column is the partial differential sensitivity R of the parameter e to the terminal model velocity Ve. (e, Ve).

Next, in S1416, the timer is reset to 0 and started, and after S1418 and S1420, S1
Proceeding to 422, it is determined whether or not there is a gait switch. If not, the process proceeds to S1432. If affirmative, the process proceeds to S1428 via S1424 and S1426 to mix gaits.

The gait mixing process will be described with reference to the flowchart of FIG. 38. In S1500, a mixed gait parameter is generated by any of the methods of the previous embodiment. At this time, a standard gait close to the initial and terminal body position / velocity demand values is selected as a base gait. In the process of generating other mixed gait parameters, the initial and terminal body position / velocity are directly calculated. The gait parameters are excluded from the setting parameters, and the mixed gait parameters are obtained as in the other embodiments. Then, the process advances to S1502, for requesting the initial and terminal body position / velocity values and the initial / terminal body position / mixed gait.
The differences Δx0, Δv0, ΔXe, ΔVe from the speed are obtained.

Subsequently, the flow advances to S1504, where the initial position / velocity of the model is set to Δx0, Δv0, and the end positions / velocities ΔXse, ΔVse when ZMP is set to 0 are obtained by the following equations. ΔXse = R (Xe, x0) * Δx0 + R (Xe, v0) * Δv0 ΔVse = R (Ve, x0) * Δx0 + R (Ve, v0) * Δv0

Then, the flow advances to S1506, where the correction amounts Δc and Δe of the ZMP parameters are obtained by the following equations. Δc = R (c, Xe) * (Δx + ΔXse) + R (c, Ve) *
(Δv + ΔVse) Δe = R (e, Xe) * (Δx + ΔXse) + R (e, Ve) *
(Δv + ΔVse)

Next, the flow advances to S1508, where the ZMP parameters c and e are obtained by the same formula as used in S1208. Subsequently, S1430 and S1 in the flow chart of FIG.
The process proceeds to the flow chart of FIG.
In steps S1602 and S1602, the same processing as in the fourth embodiment is performed, and the flow advances to step S1604 to obtain a ZMP trajectory correction amount.
This is because the parameters c = Δc and e = Δ in FIG.
e, a pattern in which the ZMP values of other breakpoints are 0.

Then, the flow advances to S1606 to set the model initial position / velocity to xmdl (0) = Δx0, vmdl (0) = Δv0, and to set ZMPmdl
, The ZMP trajectory correction amount is input, and the behavior of the model is calculated by sequential calculation. That is, the position / speed of the model at this time (time t) is obtained by a discretized model formula based on the position / speed of the model at the previous time (time t-Δt) and the ZMP trajectory correction amount of the current time (time t). However, in the first control cycle, the current position / velocity of the model is expressed as xmdl (0) = Δx
0, vmdl (0) = Δv0. Note that a matrix expression model generally used in modern control theory may be used as the discretization model.

Next, the flow advances to S1608, and according to the following equation,
The target body position is obtained by adding the model position to the mixed body position as a correction amount. x (k) = xmix (k) + xmdl (kΔt) where x (k): target body position xmix (k): k-th data of body trajectory time series of mixed gait xmdl (kΔt): time kΔt Model position.

Since the fifth embodiment is configured as described above, the same effects as those of the fourth embodiment can be obtained.

More specifically, since the linear inverted pendulum model is used, the capacity of the memory (ROM 84) is small, and it is possible to calculate the perturbation of the body trajectory in real time with a small amount of calculation by sequential calculation or the like. it can. In addition, an exact solution is easily obtained from this model formula, and future behavior can be easily predicted from the exact solution.

The sensitivity of the terminal body position / velocity to the ZMP parameter can be obtained by an exact solution. However, since an exponential function operation is included, it is more necessary to obtain and store the values in advance to further increase the operation time. Can save money.

FIG. 40 is a flow chart showing a sixth embodiment of the present invention.

As mentioned earlier, if the inertial force due to the perturbation of the initial free leg foot position and the last free leg foot position is negligible, that is, the ZMP hardly changes, the gait will be described. Can be simplified. The sixth embodiment is intended for that.

In the gait defined in this specification, the end time is the moment when the free leg lands, and at that time, the ZMP is still set on the grounding surface of the supporting leg foot. Therefore, if the inertial force due to the change of the end free leg foot position can be neglected, there is no need to change the ZMP parameter, but the initial position of the set ZMP pattern is the initial free leg foot position / posture toe. Since it is normally set to be present, when the initial free leg foot position is changed, the ZMP pattern must also be changed.

Hereinafter, ZM such as changing the initial free leg foot position
The case where the P parameter is changed is described as an example, focusing on differences from the fifth embodiment.

S170 in the flow chart of FIG.
From 0 to S1710, the process proceeds to S1712 through the same processing as in the fifth embodiment. In the sixth embodiment, in order to satisfy the dynamic equilibrium condition, the following processing is additionally changed with the change of the ZMP pattern.

That is, in S1712, as shown in FIG. 41, the value of ZMP at time 0 is set to p,
Among the ZMP parameters, c, e, and p are defined as c = 0, e =
Assuming that 0, p = 1, the value of ZMP at other break points is 0, the initial state is x0 = 0, v0 = 0, the behavior of the perturbation dynamic model is calculated until the end time of the gait, and the end model position Xe And the terminal model speed Ve.

Next, the partial differential sensitivity of the terminal model position Xe with respect to the ZMP parameter p is obtained by the following equation. R (Xe, p) = Xe

Similarly, the partial differential sensitivity of the terminal model speed Ve with respect to the ZMP parameter p is obtained by the following equation. R (Ve, p) = Ve

Subsequently, the same processing as in the fifth embodiment is performed until S1728, and the flow advances to S1730 to mix gaits. That is, in S1800 of the flow chart of FIG. 42, a mixed gait parameter is created by any method of the previous embodiment. At this time, an initial body position / velocity, a terminal body position / velocity, an initial free leg foot position, and a terminal free leg foot position that are close to required values are selected as base gaits.

However, after the base gait is selected, the initial body position / velocity, the terminal body position / velocity, the initial free leg foot position and the last free leg foot position are excluded from the directly set parameters. A mixed gait is created as in the other embodiments. Therefore, the initial body position / velocity, the end body position / velocity, the initial free leg foot position, and the last free leg foot position are values as they are in the base gait.

Subsequently, the flow advances to S1806 via the processing of S1802 and S1804, and the difference between the required initial free leg foot position and the initial free leg foot position of the mixed gait is determined, and is set to Δp. Then, the process proceeds to S1808, where the initial ZMP of the model is set to Δ
p, and the end position / velocity ΔXpe, ΔVpe when ZMP at the other breakpoints is set to 0 by the following equation. ΔXpe = R (Xe, p) * Δp ΔVpe = R (Ve, p) * Δp

Subsequently, the flow advances to S1810, where the correction amounts Δc and Δe of the ZMP parameters are obtained by the following equations. Δc = R (c, Xe) * (ΔXe + ΔXse + ΔXpe) + R
(c, Ve) * (ΔVe + ΔVse + ΔVpe) Δe = R (e, Xe) * (ΔXe + ΔXse + ΔXpe) + R
(e, Ve) * (ΔVe + ΔVse + ΔVpe)

Subsequently, returning to the flow chart of FIG.
Proceeding to 1732, the parameters of the mixed gait are substituted for the parameters of the desired gait. However, the ZMP parameter c,
The value obtained in S1812 is substituted for e. Also, a required value is substituted for the initial free leg foot position parameter and the last free leg foot position parameter.

Subsequently, the flow advances from S1734 to the flow chart of FIG. 43, and the same processing as in the fifth embodiment is performed in S1900 and S1902. As a result, a foot trajectory satisfying the requirements of the initial free leg foot position and the last free leg foot position, and a corrected ZMP trajectory are obtained. Next, the process proceeds to S1904, where a ZMP trajectory correction amount is obtained. As shown in FIG. 44, this is a ZMP pattern in which the parameters c = Δc, e = Δe, p = Δp, and the ZMP values of other break points are set to 0.

The remaining structure is not different from that of the fifth embodiment.

Since the sixth embodiment is configured as described above, it is possible to obtain the same effects as those of the fourth to fifth embodiments and to use a standard gait with different initial body position and speed. Since there is no need to store the information as a storage, the storage capacity can be further reduced.

FIG. 45 is a block diagram of a gait generator similar to that of the first embodiment shown in FIG. 3, showing a seventh embodiment of the present invention. However, the method of calculating the body horizontal position is different in the seventh embodiment.

In the seventh embodiment, the positions and postures of both feet of the obtained mixed gait at this time (time t), ZMP
In addition, the body height Z and the previous state (time t-Δt) (body position / velocity, etc.) are represented by a robot motion model (including dynamics, kinematics, various constraint equations, and constraint equation).
The current value (instantaneous value) of the horizontal coordinate (X, Y) of the body position satisfying the dynamic equilibrium condition is calculated.

That is, in the seventh embodiment, the second
As in the embodiment of FIG.
000 to 2014 to S in the flow chart of FIG.
Returning to FIG. 46 through the processing from 2100 to S2104,
The process proceeds from S2022 to S2200 and subsequent steps in FIG.
06, the upper body of the mixed gait such that the ZMP dynamically calculated from the motion of the upper body of the mixed gait and the foot of the mixed gait matches the set ZMP obtained from the ZMP parameters of the mixed gait. The trajectory was determined by sequential calculation.

However, in the initial stage of the gait, appropriately set initial body position / speed parameter values are used for the body position / speed. In other words, the gait of one step is obtained by sequentially calculating by sequential calculation, starting from the initial body position / velocity parameter values.

This method has the disadvantage that the amount of calculation is large and the end body position / posture slightly deviates from the required value, but the dynamic equilibrium condition can be strictly satisfied.

In the seventh embodiment, the body height and the body horizontal position at this time may be determined so as to simultaneously satisfy the dynamic equilibrium condition and the constraint equation of the body height determination method.
In the previous embodiment, in order to reduce the calculation time, the body height was calculated using the horizontal position of the body last time, assuming that the body height does not greatly change in a short time.

As described above, in the first to seventh embodiments, at least the upper body 24 and the plurality of leg links connected to the upper body via the joints 10, 12, 14R (L). 2. A gait generator for a legged mobile robot comprising: a standard gait storage for storing a plurality of types of standard gaits each comprising a set of parameters including a parameter relating to a floor reaction force for at least one gait; Means (such as S10), gait requesting means (such as S24) for making a request relating to a gait, and selecting one or more of the standard gaits in response to the request relating to the gait. Gait generating means (S26 to S30, S100 to S106, S2) for generating an approximate gait satisfying the requirement for the gait by performing an approximate calculation based on
00 to S206).

Further, the gait generator is configured to obtain the approximate gait by performing a weighted average of the selected gaits (S100 to S106, S200 to S206, etc.).

Also, the gait generator is configured to obtain the approximate gait using the parameter sensitivity of the selected gait (S500, S600 to S604, S700 to S706, etc.).

The gait generator averages the weight of the selected gait and obtains the approximate gait using the parameter sensitivity of the selected gait (S800, S800).
900 to S906, S1000 to S1006, etc.).

Further, the gait generating means obtains and stores, as the parameter sensitivity of the gait, a change amount of another parameter per unit change amount of at least one of the standard gaits. S800, etc.), and the approximate gait is obtained using the stored parameter sensitivity of the gait.

The standard gait generating means includes expression means (S1104 to S1114, etc.) for expressing the relationship between the parameters relating to the floor reaction force of the standard gait and other parameters. The gait is corrected based on the relationship (1), and a gait that satisfies the requirement for the gait is generated such that at least the displacement and the speed are continuous in the boundary condition (S1128, S1128).
1200 to S1208).

In addition, the perturbation of the other parameters is configured to be a perturbation of the horizontal position of the body.

[0243] Further, the relationship is configured to be expressed by a linear model.

Further, the above-mentioned relationship is configured to be expressed by an inverted pendulum model.

Also, the relationship is configured to be represented by a time-series value.

When the parameter relating to the floor reaction force is under a predetermined condition, the gait generating means obtains the approximate gait by excluding the parameter (S1730,
S1800 to S1812).

The gait generating means calculates joint angles of the robot based on the generated gaits (S302), and the gait generating means calculates the joint angles of the robot so as to obtain the calculated joint angles. It is configured to include a joint drive control unit (S306) for driving and controlling the joint.

The joint angle determining means detects the state quantity of the robot (S300, S400
To S406), and correcting means (S) for correcting the position and orientation of the robot based on the detected state quantity.
408), wherein the joint angle calculating means calculates the joint angle of the robot so as to obtain the corrected position and posture.

In the first to seventh embodiments, examples of approximation operations are not limited to those shown in the drawings, and various other modifications are possible.

In the first to seventh embodiments, the body position is described by coordinates of a certain reference point in the body 24 as viewed from the support leg contact point. The reference point should be set at the center in the left-right direction, but there are several setting methods for the front-rear direction as follows. Setting method 1) Set to the center of gravity of the body (adopted in the first to seventh embodiments) Setting method 2) Set to a position that coincides with the overall center of gravity position when standing upright. Setting method 3) Set to an appropriate position by trial and error.

No matter which setting method is used, there is no difference in the case of walking in which the direction of the body is constant, but in the case where the direction of the body changes, such as in turning walking, the reference point is not changed. The manner in which the error occurs differs depending on the method of taking. For example, as shown in FIG. 49, consider the case where the upper body is rotated around the reference point while moving the reference point rightward on the paper.
It can be seen that the positional relationship of the upper body is clearly different between the case where the reference point is at the center of the upper body and the case where the reference point is closer to the front.

Although an example has been described in which gaits having different turning angles are mixed when generating a mixed gait (required gait), mixing gaits is a kind of approximation. However, for the above reasons, the mixed gaits are slightly different depending on the setting of the reference point, and the approximation accuracy for the dynamic equilibrium condition changes.

Therefore, the reference point should be set so as to minimize the error with respect to the dynamic equilibrium condition. Which of the above setting methods is optimal or which does not cause any problem depends on the structure of the robot and the way of walking.

Further, in order to improve the accuracy of the dynamic equilibrium condition, a time series table of the center of gravity of the entire robot may be used instead of the time series table of the body position. However, as a disadvantage, considerable computation is required to determine the posture of the robot so that the center of gravity of the entire robot is at the designated position.

In the above description, (displacement) acceleration has not been described as a gait parameter, but it goes without saying that a smoother gait is generated when the acceleration is adjusted at the boundary of the gait.

Although the present invention has been described with respect to a bipedal walking robot, the present invention can be applied not only to a bipedal walking robot but also to a multilegged robot.

[0257]

As described above, a gait including a floor reaction force can be freely and in real time generated to realize an arbitrary stride, turning angle, and the like. Further, the displacement and the speed of each part of the robot can be made continuous at the boundary between the generated gaits. Further, it is possible to drive and control the legged mobile robot based on the gait generated in real time.

[Brief description of the drawings]

FIG. 1 is an explanatory view showing an entire gait generating device for a legged mobile robot according to the present invention.

FIG. 2 is a block diagram showing details of a control unit of the bipedal walking robot shown in FIG. 1;

FIG. 3 is a block diagram functionally showing the operation of the gait generator of the legged mobile robot according to the present invention.

FIG. 4 is a main flowchart showing the operation of the gait generator of the legged mobile robot according to the present invention.

FIG. 5 is an explanatory diagram showing a free leg position and posture of a zeroth standard gait in a standard gait used in the gait generation operation of FIGS. 3 and 4;

6 is a timing chart showing a foot trajectory of the 0th standard gait of FIG. 5;

FIG. 7 is a timing chart of ZMP (X coordinate) of the 0th standard gait of FIG. 5;

FIG. 8 is an explanatory diagram showing a free leg position and a posture of a first standard gait among standard gaits used in the gait generation operation of FIGS. 3 and 4;

9 is a timing chart of ZMP (X coordinate) of the first standard gait of FIG. 8;

FIG. 10 is an explanatory diagram showing a free leg position and a posture of a second standard gait among standard gaits used in the gait generation operation of FIGS. 3 and 4;

11 is a timing chart of ZMP (X coordinate) of the second standard gait of FIG. 10;

FIG. 12 is an explanatory diagram showing a free leg position and posture of a required gait requested in the gait generating operation of FIGS. 3 and 4.

13 is a timing chart of ZMP (X coordinate) of the required gait of FIG.

FIG. 14 is a subroutine flowchart showing a gait mixing operation in the flowchart of FIG. 4;

FIG. 15 is an explanatory diagram for explaining the operation of the flow chart of FIG. 14;

FIG. 16 is an explanatory diagram illustrating the operation of the flow chart of FIG. 14 similarly to FIG. 15;

FIG. 17 is a subroutine flowchart showing an operation of calculating an instantaneous value of a desired gait in the flowchart of FIG. 4;

FIG. 18 is a flow chart of double leg compliance control proposed by the present applicant based on a gait generated by the gait generation operation of FIGS. 3 and 4.

FIG. 19 is a subroutine flow chart showing the operation of calculating a two-leg compliance operation amount in the flow chart of FIG. 18;
It is a chart.

FIG. 20 is a main flow chart similar to FIG. 4, showing the operation of the device according to the second embodiment of the present invention.

FIG. 21 is a subroutine flowchart showing the gait mixing work in the flowchart of FIG. 20;

FIG. 22 is a subroutine flowchart showing an operation of calculating an instantaneous value of a desired gait in the flowchart of FIG. 20;

FIG. 23 is a main flow chart similar to FIG. 4, showing the operation of the device according to the third embodiment of the present invention.

FIG. 24 is a subroutine flowchart showing a gait mixing operation in the flowchart of FIG. 23;

FIG. 25 is a subroutine flowchart showing an operation of calculating an instantaneous value of a desired gait in the flowchart of FIG. 23;

FIG. 26 is a main flow chart similar to FIG. 4, showing the operation of the device according to the fourth embodiment of the present invention.

FIG. 27 is a timing chart showing a gait such as a body position generated in the previous embodiment.

FIG. 28 is a timing chart of ZMP (X coordinate) for explaining boundary condition adjustment in the operation of the flow chart of FIG. 26;
It is a chart.

FIG. 29 is a ZMP (X coordinate) timing chart for explaining the operation of the operation in the flow chart of FIG. 26;

30 is a timing chart of ZMP (X coordinate) similarly illustrating the operation of the operation in the flowchart of FIG. 26.

FIG. 31 is a timing chart of ZMP (X coordinate) similarly illustrating the operation of the operation of the flowchart of FIG. 26.

32 is a timing chart of ZMP (X coordinate) similarly illustrating the operation of the operation of the flow chart of FIG. 26.

FIG. 33 is a subroutine flowchart showing a gait mixing operation in the flowchart of FIG. 26;

FIG. 34 is a subroutine flowchart showing an operation of calculating an instantaneous value of a desired gait in the flowchart of FIG. 26;

FIG. 35 is an explanatory timing chart showing gaits generated in the fourth embodiment shown in the flow chart of FIG. 26;

FIG. 36 is a main flow chart similar to FIG. 4, showing the operation of the device according to the fifth embodiment of the present invention.

FIG. 37 is an explanatory diagram of an inverted pendulum model used in the fifth embodiment.

FIG. 38 is a subroutine flowchart showing the gait mixing operation in the flowchart of FIG. 36;

FIG. 39 is a subroutine flowchart showing an operation of calculating an instantaneous value of a desired gait in the flowchart of FIG. 36;

FIG. 40 is a main flow chart similar to FIG. 4, showing the operation of the device according to the sixth embodiment of the present invention.

FIG. 41 is a timing chart for explaining the operation of the fifth embodiment in the flow chart of FIG. 40;

FIG. 42 is a subroutine flowchart showing the gait mixing operation in the flowchart of FIG. 40;

FIG. 43 is a subroutine flowchart showing an operation of calculating an instantaneous value of a desired gait in the flowchart of FIG. 40;

FIG. 44 is a timing chart illustrating the operation of the flow chart of FIG. 40;

FIG. 45 is a block diagram similar to FIG. 3, showing the configuration and operation of a device according to a seventh embodiment of the present invention.

FIG. 46 is a main flow chart showing the operation of the device according to the seventh embodiment.

FIG. 47 is a subroutine flowchart showing a gait mixing operation in the flowchart of FIG. 44;

FIG. 48 is a subroutine flowchart showing an operation of calculating an instantaneous value of a desired gait in the flowchart of FIG. 44;

FIG. 49 is an explanatory diagram illustrating the position of the upper body in the gait generated by the gait generating device of the legged mobile robot according to the present invention.

[Explanation of symbols]

Reference Signs List 1 bipedal walking robot (legged mobile robot) 2 leg link 10, 12, 14R, L hip joint 16R, L knee joint 18, 20R, L ankle joint 22R, L foot 24 upper body 26 control unit

Claims (13)

[Claims] 1. A gait generator for a legged mobile robot comprising at least an upper body and a plurality of leg links connected to the upper body via joints, comprising: a. Standard gait storage means for storing a plurality of types of standard gaits each comprising a set of parameters including parameters relating to floor reaction force for at least one gait; b. Gait request means for making a request regarding a gait; and c. An approximate gait that satisfies the gait requirement is selected by selecting one or more of the standard gaits in response to the gait requirement and performing an approximate operation based on the selected standard gait. A gait generator for a legged mobile robot, comprising: 2. The gait generator for a legged mobile robot according to claim 1, wherein said gait generator obtains the approximate gait by performing a weighted average of the selected gait. 3. The gait generating device for a legged mobile robot according to claim 1, wherein said gait generating means obtains the approximate gait using parameter sensitivity of the selected gait. 4. The gait generator according to claim 1, wherein the gait generator averages the weight of the selected gait and obtains the approximate gait using parameter sensitivity of the selected gait. Gait generator for legged mobile robots. 5. The gait generator includes: d. Storage means for obtaining and storing a change amount of another parameter per unit change amount of at least one of the standard gaits as a parameter sensitivity of the gait, and using the stored parameter sensitivity of the gait; 5. The gait generator for a legged mobile robot according to claim 3, wherein the approximate gait is obtained by calculating the gait. 6. The standard gait generator includes: e. Expression means for expressing a relationship between perturbations of other parameters with respect to parameters relating to the floor reaction force of the standard gait,
Comprising, correcting the approximate gait based on the expressed perturbation relationship, generating a gait that satisfies the gait requirement so that at least the displacement and velocity are continuous in the boundary condition. A gait generator for a legged mobile robot according to any one of claims 1 to 5. 7. The gait generator for a legged mobile robot according to claim 6, wherein the perturbation of the other parameter is a perturbation of a horizontal position of the upper body. 8. The gait generator for a legged mobile robot according to claim 6, wherein the relation is represented by a linear model. 9. The gait generator for a legged mobile robot according to claim 8, wherein the relationship is represented by an inverted pendulum model. 10. The gait generator for a legged mobile robot according to claim 6, wherein the relation is expressed by a time-series value. 11. The gait generator according to claim 1, wherein when the parameter relating to the floor reaction force is under a predetermined condition, the gait generator excludes the parameter and obtains the approximate gait. Item. The legged mobile robot gait generator according to any one of the above items. 12. The gait generator includes: f. Joint angle calculating means for calculating a joint angle of the robot based on the generated gait; and g. The gait generation of the legged mobile robot according to any one of claims 1 to 11, further comprising: an articulation drive control unit that drives and controls a joint of the robot so that the calculated joint angle is obtained. apparatus. 13. The joint angle determining means: h. Detecting means for detecting a state quantity of the robot; and i. Correction means for correcting the position and orientation of the robot based on the detected state quantity, wherein the joint angle calculation means calculates the joint angle of the robot so as to be the corrected position and attitude. 13. The gait generator for a legged mobile robot according to claim 12, wherein