WO2003068453A1

WO2003068453A1 - Two-leg walking moving device

Info

Publication number: WO2003068453A1
Application number: PCT/JP2003/001325
Authority: WO
Inventors: Takayuki Furuta; Tetsuo Tawara; Yu Okumura; Hiroaki Kitano
Original assignee: Japan Science And Technology Corporation
Priority date: 2002-02-18
Filing date: 2003-02-07
Publication date: 2003-08-21
Also published as: JP2003236783A; TW200303257A

Abstract

A two-leg walking moving device comprising join-use drive motors for respectively swinging leg-swinging joints (15L, 15R) of the waist, rolling-direction joints (16L, 16R) of the waist, pitch-direction joints (17L, 17R) of the waist, pitch-direction joints (18L, 18R) of knees (12L, 12R), pitch-direction joints (19L, 19R) of ankles with respect to feet (14L, 14R), and roll-direction joints (20L, 20R) of ankles, and a walking control device (30) for producing gait data at a gait generation unit (31) in response to a motion command and drive-controlling respective drive motors at a control unit (32) based on this gait data, wherein the gait generation unit (31) detects a robot’s attitude based on sensor information from an angle detection sensor (24) for detecting the walking conditions of respective joints and from ZMP detection sensors (23L, 23R) for detecting ZMP at each foot, produces in real time angle command values as final gait data from a target value by a motion command and the current values of sensor information based on this attitude, acquires environmental information for each walk, and controls a walk in real time according to this environmental information.

Description

Description Bipedal locomotion device Technical field

The present invention relates to a biped walking type mobile device that performs biped walking such as a biped walking humanoid robot, and actively acquires environmental information such as a road surface to be walked, and responds to this environmental information. The present invention relates to a biped walking type mobile device capable of performing controlled walking. Background art

Conventionally, a humanoid robot as a so-called bipedal locomotion device generates gait patterns (hereinafter referred to as “gait”) set in advance and performs gait control in accordance with the gait data. Predetermined walking no. By walking the legs in turns, biped walking is realized. At this time, in order to stabilize the walking posture, the point at which the combined moment of the floor reaction force and gravity at the sole of the robot becomes zero (hereinafter referred to as ZMP (Zero Moment Point)) is set as the target value. The robot is stabilized according to this ZMP standard by performing convergence, so-called ZMP compensation.

By the way, in a conventional bipedal walking humanoid robot, a bipedal walking of a robot is realized by a single walking control method on the premise that the robot walks on a road surface with known environmental information such as a flat road surface. Has become. Therefore, under such preconditions, irregular terrain such as unevenness on the road surface is treated as a disturbance in walking control, and a compensation unit is provided separately from the gait data generation unit. The part absorbs and removes disturbance.

However, there is a limit in dealing with unknown external tongue elements using only the compensator, and the stability of robot bipedal walking may be impaired. As for the walking control method, it is difficult to stabilize walking by using only a single walking control method, for example, to respond promptly to environmental changes such as road surface conditions.

Actively acquire and recognize environmental information such as the posture of the mouth bot and the road surface that should be walked. Thus, it is possible to stabilize walking by creating an exercise plan in real time, but such walking control is not performed in a conventional bipedal walking robot. Disclosure of the invention

In view of the above, the present invention provides a bipedal locomotion device which acquires environmental information for each walk and performs walking control in real time in response to the environmental information. The purpose is.

According to a first aspect of the present invention, there is provided a main body, a leg having a knee at an intermediate position on both lower sides of the main body and capable of swinging in two axial directions, and a leg at a lower end. Drive means for swinging the swingable joints of the legs, lower legs, and thighs, respectively, and a gait generator corresponding to the target angle trajectory, target angular velocity, and target angle corresponding to the movement command. A walking control device that generates gait data including acceleration, and drives and controls each of the driving units by a control unit based on the gait data. The gait generator generates the robot based on the sensor information from the sensors for detecting the state of each joint during walking and / or the sensor for detecting the angular velocity and angular acceleration and the foot force sensor mounted on the robot. Detects the posture and based on this posture Ri by the current value from biped walking mobile apparatus characterized by generating in real time an angle command value of the final gait de one data target value and the sensor information by the movement instruction is achieved.

In the two-legged walking type mobile device according to the present invention, preferably, the gait generator generates a gait library storing gait modules serving as elements of robot walking, and gait data relating to the next gait. A gait module that selects and reads the corresponding gait module from the gait library, and a compensator that generates the final gait data by synthesizing the gait module selected by the selector in real time. And

In the two-legged walking device according to the present invention, preferably, the gait module stored in the gait library is constituted by at least one angular trajectory pattern.

The bipedal locomotion device according to the present invention preferably comprises the gait module The angular trajectory pattern is classified as ZMP value, angular momentum around the center of gravity, and angular acceleration as indices.

In the bipedal walking type moving device according to the present invention, preferably, the selection unit detects the robot posture from the sensor information after the end of the previous walking, and sets a position until the target end state of the next walking by the movement command. A motion plan is made, and a gait module necessary for realizing the motion plan is read from the gait library.

According to a second aspect of the present invention, there is provided a main body, a leg having a knee at a middle portion capable of swinging in two axial directions on both lower sides of the main body and a leg at a lower end; A gait generating unit corresponding to a motion command is provided for a bipedal walking type moving device including a driving means for swinging a swingable joint of a leg, a lower leg, and a thigh of the leg. A gait data including a target angle trajectory, a target angular velocity, and a target angular acceleration is generated by using the gait data, and based on the gait data, a control unit drives and controls each of the above driving means by a control unit. The gait generator of the walking control device may include a sensor for detecting a state of each of the joints during walking and / or a sensor for detecting an angular velocity and an angular acceleration mounted on a lopot, and a sensor from a foot force sensor. The attitude of the robot based on the information And a real-time angle command value as final gait data is generated from a target value based on the motion command and the current value of the sensor information based on the posture. Achieved.

In the walking control device for a bipedal walking type moving device according to the present invention, preferably, the gait generator stores a gait module storing a gait module that is an element of robot walking; When generating gait data, the gait module that selects and reads the corresponding gait module from the gait library and the gait module selected by the selection unit are synthesized in real time to generate final gait data And a compensating unit.

The gait module stored in the gait library preferably comprises at least one angular trajectory pattern.

In the walking control device for a bipedal walking type moving device according to the present invention, preferably, the angular trajectory pattern constituting the gait module includes a ZMP value, an angular momentum around a center of gravity, an angular addition. Speed is classified as an index.

In the walking control device for a bipedal walking type mobile device according to the present invention, preferably, the selection unit detects a posture of the robot from sensor information after the end of the previous walking, and sets a target of a next walking by a movement command. A motion plan up to the terminal state is made, and a gait module necessary for realizing the motion plan is read from the gait library.

According to the above configuration, when the biped walking type mobile device performs a walking motion, the gait generator of the walking control device performs the robot control based on the sensor information from the sensors provided in the robot when the robot walks. Then, an angle command value for the final gait is generated based on this posture. Therefore, the gait generator always generates gait data in real time based on the posture of the robot during the walking motion of the robot, so that the walking motion can be stably and reliably performed even on a road surface on which environmental information is unknown. Will be able to do so.

The gait generator selects a gait library that stores gait modules that are elements of robot gait and a corresponding gait module from the gait library when generating gait data for the next walk When the gait generator includes a selecting unit that reads out the gait module selected by the selecting unit and a compensating unit that generates final gait data by synthesizing the gait module in real time, When generating gait data, desired gait data can be generated by reading gait modules as elements from the gait library and combining them. As a result, the amount of calculation by the gait generator is reduced, and gait data can be generated quickly.

If the gait modules stored in the gait library are composed of at least one angular orbital pattern, more diverse gait data can be synthesized.

If the angular trajectory patterns that make up the self-gait module are classified using the ZMP value, angular momentum around the center of gravity, and angular acceleration as indices, the gait generator selects the gait module from the library. When reading, the gait module can be quickly searched and read.

After the previous walking is completed, the selecting unit detects the robot posture from the sensor information, and performs an operation plan up to a target end state of the next walking based on the movement command, and When reading out the gait modules necessary to realize the work plan from the gait library, the selection unit performs an operation plan for the next walk based on the sensor information for each walk, and the corresponding gait module. Select a module. Therefore, the compensator can generate gait data in real time based on the gait module selected by the selector for each walk. BRIEF DESCRIPTION OF THE FIGURES

FIG. 1 is a schematic diagram showing a mechanical configuration of an embodiment of a bipedal walking robot according to the present invention.

FIG. 2 is a block diagram showing an electric configuration of the biped walking robot of FIG. FIG. 3 is a block diagram showing a configuration of the walking control device in the bipedal walking robot of FIG.

FIG. 4 is a flowchart showing walking control of the biped walking robot of FIG. BEST MODE FOR CARRYING OUT THE INVENTION

Hereinafter, the present invention will be described in detail based on an embodiment shown in the drawings. 1 and 2 show the configuration of an embodiment of a bipedal walking robot to which the bipedal walking type moving device according to the present invention is applied.

In FIG. 1, a bipedal walking robot 10 has an upper body 11 as a main body, and two legs with knees 1 L and 1 2 R in the middle attached to both lower sides of the upper body 11. 13 L, 131¾, and legs 13 L, 14 L, 14 R attached to the lower end of the 13 R.

Here, the upper legs 13 L and 13 R are respectively six joints, that is, joints 15 L and 15 for turning the waist legs relative to the upper body 11 in order from the top. R, joints in the waist roll direction (around the X axis) 16 L, 16 R, joints in the hip pitch direction (around the y axis) 17 L, 171¾, knees 12 and 1 2 R Pitch direction joints 18L, 18R, feet 14L, 14R, ankle pitch joints 19L, 19R, ankle roll direction joints 20L, 20 It has R. Each of the joints 15L, 15R to 20L, 20R is constituted by a joint driving motor. this Thus, the waist joint is composed of the above-mentioned joints 15 and 15 R, 16 R, 16 L, 16 R, 17 L and 17 R, and the ankle joint is the joints 19 L and 1 It consists of 9 R, 20 L, and 20 R. Further, the waist joint and the knee joint are connected by thigh links 21 L and 21 R, and the knee joint and the ankle joint are connected by a crus link 22 L,

They are linked by 22R.

As a result, the legs 13 L, 13 R and the feet 足 IH 4 L, 14 R on both the left and right sides of the bipedal walking robot 10 are given 6 degrees of freedom, respectively. By controlling the drive of these 12 joints at appropriate angles, angular velocities, and angular accelerations by drive motors, the legs 13 L, 13 R, and the legs 14 L, 14 R are all By giving a desired motion, the user can walk in a three-dimensional space arbitrarily.

Further, the above-mentioned feet 14 L and 14 R are ZMP detection sensors 2 as foot force sensors.

3 L, 23 R are provided. The ZMP detection sensors 23 L and 23 R detect the ZMP that is the center point of the sole reaction force of the soles 14 L and 14 R, respectively, and output the measured ZMP value.

Further, each joint is provided with an angle detection sensor 24 (described later) such as a rotary encoder corresponding to the driving means.

The upper body 11 is simply shown as a box T in the illustrated case, but may actually have a head or both hands.

FIG. 2 shows the electrical configuration of the bipedal walking robot 10 shown in FIG. In FIG. 1, a bipedal walking robot 10 includes a walking control device that drives and controls the above-mentioned driving means, that is, the above-described joint driving motors 15 L, 15 R to 20 L, and 20 R of each joint. 30. The coordinate system of the bipedal walking robot 10 is an x-yz coordinate system with the X-direction (forward +) in the front-rear direction, the y-direction (inward +) in the lateral direction, and the z-direction (up +) in the up-down direction. Use

The gait control device 30 includes a gait generator 31 that generates gait data in response to a motion command, and a driving unit based on the gait data. And a control unit 32 for driving and controlling the motors 15L, 15R to 20L, 20R.

The gait generator 31 is a bipedal walking type corresponding to a motion command input from the outside. Each joint 15 required for the robot 10 to walk 15 generates gait data including the target angular trajectory, target angular velocity, and target angular acceleration of 15R to 20L and 20R. . Here, the gait generator 31 includes a selector 33 and a compensator 34, as shown in FIG. '

After the previous walking is completed, that is, after the swing leg impact shock absorption period is completed, the selection unit 33 receives the sensor information from the angle detection sensor 24 and the ZMP detection unit 23, 3R, and the angular velocity or angular velocity. The robot's posture is detected by acquiring the angular value of each joint of the robot, the angular momentum around the center of gravity, the angular acceleration, and the ZMP value from the acceleration detection sensor and the sole force sensor.

Here, the selection unit 33 includes a gait library 33a. The gait library 33a previously stores a gait module 33b as posture data which is an element of the robot's walking motion. The gait module 33b is composed of, for example, at least one angular trajectory pattern. Each angular trajectory pattern is categorized as ZMP value as sensor information, angular momentum around the robot's center of gravity, and angular acceleration as intex. Then, the selection unit 33 performs an operation plan by calculating the initial posture, the terminal posture, and the initial angular momentum required for the movement of the robot in the single-leg supporting period, and according to the motion plan, The gait module 33b, which provides the movement in the two-leg support period necessary to realize the initial posture during the support period, is retrieved from the gait library 33a, read out, and output to the compensator 34.

The compensating unit 34 combines these gait modules 33 b in real time by at least one gait module 33 b from the selecting unit 33, and calculates the final value of each joint as gait data. Generates an ideal angle command value. Further, the compensator 34 performs ZMP target value compensation of the gait data synthesized in real time based on the ZMP actual measurement values from the ZMP detection sensors 23L and 23R.

Thus, the gait generator 31 outputs the final angle command value of each joint to the controller 32 as gait data for each walk.

The control unit 32 generates a control signal for each joint driving motor based on the angle command value as gait data from the compensating unit 34 and the sensor information from the angle detection sensor 24. Drive the joint drive motor of each joint of the robot 10 according to the signal. Dynamic control.

The bipedal walking robot 10 according to the embodiment of the present invention is configured as described above, and walks as shown in the flowchart of FIG.

In the flowchart of FIG. 4, first, in step ST1, the gait generator 31 of the walking control device 30 designates a target end state of the robot for the next walking when the previous walking is completed. Then, the operation plan is started in step ST2. Then, in step ST3, the selection unit 33 of the gait generation unit 31 sends sensor information and / or angular velocity and angular acceleration from the angle detection sensor 24 and the ZMP detection sensors 23L and 23R. Based on information from the detection sensor and the sole force sensor, the current angle value, angular momentum around the center of gravity, angular velocity, and ZMP value of each joint of the robot are detected. After that, in step ST4, the selection unit 33 calculates the angular momentum of each joint of the mouth robot required to transition from the current initial state of the robot to the target end state based on the detected values. I do.

Accordingly, in step ST5, the selection unit 33 stores the detected values as indices in the gait library 33a based on the respective detected values in step ST3 and the respective momentums in step ST4, for example. The retrieved gait module 33b is searched, and at least one corresponding gait module 33b is selected and read out, or a gait is calculated and generated in real time based on the detected value. Thus, the operation plan ends at step ST6.

Subsequently, the compensator 34 starts the robot operation in step ST7, and in step ST8, the gait module 33b selected and read by the selector 33 in the motion plan. Are combined in real time to calculate the angle and ZMP target value of each joint of the robot, and in step ST9, sensor information from the angle detection sensor 24 and ZMP detection sensor 23L, 23R Based on, the actual measured ZMP value and the momentum of each joint are detected. Note that, in step ST5, when the gait is generated in real time by dynamics calculation, the above-described synthesis may not be performed. Then, in step ST10, the compensating unit 34 compares the calculated angular momentum and ZMP target value of each joint with the detected angular momentum and ZMP measured value of each joint. Here, in step ST10, if there is no error between the calculated value and the detected value, the compensator 3 4 Determines in step ST11 whether ZMP compensation is possible. If ZMP compensation is not possible, an operation plan is performed again in step ST12, and the process returns to step ST3. If ZMP compensation is possible in step ST11, the compensator 34 performs ZMP compensation in step ST13 to change the angular motion trajectory of each joint of the robot. Instead, correct the error by adjusting only the angular velocity and angular acceleration. If there is no error between the calculated value and the detected value in step ST10, and after correcting the error in step ST13, the compensator 34 returns the current robot It is determined whether or not the state is the target end state according to the operation plan. If the state is not the target end state, the process returns to step ST8 again. If it is determined in step ST14 that the vehicle is in the target terminal state, the walking control device 30 completes the robot operation in step ST15 and ends one walk. As described above, the walking control relating to one walking of the robot is completed. The walking control device 30 repeats the above operation in real time every time the robot walks.

As described above, according to the bipedal walking robot 10 according to the embodiment of the present invention, when the mouth bot is walking, the posture of the robot at that time is detected, and the walking data for realizing the movement command is obtained. Is generated in real time, so that even when walking on a road or the like where environmental information is unknown, stable biped walking can always be realized. Thus, for example, even on a road with complicated unevenness, the robot posture is always detected, and gait data is generated while referring to the robot posture. Exercise can be performed stably and reliably.

In the embodiment described above, the legs 12 L and 12 R have six degrees of freedom, and the arms 13 L and 13 R have five degrees of freedom, but are not limited thereto. It may have a degree of freedom or a greater degree of freedom. In step ST15, a gait may be generated in real time by dynamics calculation based on the sensor information. In this case, real-time synthesis of gaits does not have to be performed. Industrial applicability

As described above, according to the present invention, when the two-legged walking device performs a walking motion, the gait generator of the walking control device is provided in each part when the robot walks. The robot posture is detected based on the sensor information from the user, and an angle command value as final gait data is generated based on the posture. Therefore, the gait generator always generates gait data in real time based on the posture of the robot during the walking motion of the robot, so that the walking motion can be stably and reliably performed even on a road surface where environmental information is unknown. It is possible to do.

Thus, according to the present invention, there is provided an extremely excellent bipedal walking type moving device which acquires environmental information for each walk and performs walking control in real time in accordance with the environmental information. Provided.

Claims

The scope of the claims

1. A main body, a leg with a knee at the middle and a foot at the lower end that can swing in two axial directions on both lower sides of the main body, and a foot, a lower leg, and a thigh of the leg. The gait generator generates the gait data including the target angular trajectory, the target angular velocity, and the target angular force D speed by the gait generator in response to the motion command. A walking control device that drives and controls each of the above-described driving means by a control unit based on the gait data;

The gait generator of the walking control device is configured to detect a state of each of the joints during walking and / or sensor information from an angular velocity and angular acceleration detecting sensor mounted on the robot and a foot force sensor. The robot is configured to detect the robot's posture, and based on the posture, generate an angle command value as final gait data in real time from a target value by a motion command and a current value of sensor information. Leg walking type moving device.

2. The gait generator stores a gait module that stores gait modules that are elements of robot walking, and a gait module corresponding to the gait library when generating gait data for the next gait. 2. A selecting unit for selecting and reading the selected gait module, and a compensating unit for real-time synthesizing the gait module selected by the selecting unit to generate final gait data. Bipedal walking mobile device described in

3. The bipedal locomotion device according to claim 2, wherein the gait module stored in the gait library includes at least one angular trajectory and a 'turn'.

4. The bipedal locomotion according to claim 3, wherein the angular trajectory pattern constituting the gait module is classified as a ZMP value, an angular momentum around a center of gravity, and an angular acceleration as indices. apparatus.

5. After the previous walking is completed, the selection unit detects the posture of the robot from the sensor information, and performs an operation plan up to a target end state of the next walk by a motion command, thereby realizing the motion plan. The bipedal locomotion device according to any one of claims 2 to 4, wherein a gait module required for the purpose is read out from a gait library.

6. A main body, a leg with a knee at the middle and a foot at the lower end that can swing in two axial directions on both lower sides of the main body, and a foot, a lower leg, and a thigh of the leg. The drive means for swinging the swingable joints, respectively, and the bipedal locomotion system including the gait generator, the target angle trajectory, the target angular velocity, and the target angular acceleration are calculated by the gait generator in response to the movement command. In the walking control device of the bipedal walking type moving device that generates gait data including the gait data and drives and controls each of the driving means by the control unit based on the gait data,

The gait generator of the walking control device receives the sensor information from the sensor for detecting the state of each joint during walking and / or the sensor for detecting angular velocity and angular acceleration mounted on the robot, and the foot force sensor. Based on the posture, and based on the posture, generates an angle command value as final gait data in real time from a target value by a motion command and a current value of sensor information. A walking control device for a legged walking device.

7. The gait generation unit stores a gait library storing gait modules that are elements of robot walking, and a gait module corresponding to the gait library when generating gait data related to the next gait. A gait module selected by the selector and a compensator for generating final gait data by synthesizing the gait module selected in real time. Item 7. The walking control device for a bipedal locomotion device according to Item 6.

8. The gait module of claim 7, wherein the gait module stored in the gait library comprises at least one angular trajectory, a 'turn'. Control device.

9. The bipedal locomotion according to claim 8, wherein the angle flit pattern constituting the gait module is classified as ZMP value, angular momentum around the center of gravity, and angular acceleration as indettas. The walking control device of the device.

10. The selection unit detects the posture of the robot from the sensor information after the end of the previous walk, and performs an operation plan up to a target end state of the next walk based on the motion command, thereby realizing the motion plan. 10. The walking control device for a bipedal locomotion device according to claim 7, wherein a gait module necessary for the reading is read from a gait library.