WO2021095471A1 - Robot marcheur bipède et procédé de commande de robot marcheur bipède - Google Patents

Robot marcheur bipède et procédé de commande de robot marcheur bipède Download PDF

Info

Publication number
WO2021095471A1
WO2021095471A1 PCT/JP2020/039605 JP2020039605W WO2021095471A1 WO 2021095471 A1 WO2021095471 A1 WO 2021095471A1 JP 2020039605 W JP2020039605 W JP 2020039605W WO 2021095471 A1 WO2021095471 A1 WO 2021095471A1
Authority
WO
WIPO (PCT)
Prior art keywords
com
control
legs
gravity
equation
Prior art date
Application number
PCT/JP2020/039605
Other languages
English (en)
Japanese (ja)
Inventor
春江 付
Original Assignee
本田技研工業株式会社
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 本田技研工業株式会社 filed Critical 本田技研工業株式会社
Publication of WO2021095471A1 publication Critical patent/WO2021095471A1/fr

Links

Images

Classifications

    • B—PERFORMING OPERATIONS; TRANSPORTING
    • B25—HAND TOOLS; PORTABLE POWER-DRIVEN TOOLS; MANIPULATORS
    • B25J—MANIPULATORS; CHAMBERS PROVIDED WITH MANIPULATION DEVICES
    • B25J5/00—Manipulators mounted on wheels or on carriages

Definitions

  • the present invention relates to a bipedal walking robot and a control method for the bipedal walking robot.
  • the present application claims priority based on Japanese Patent Application No. 2019-204046 filed on November 11, 2019, the contents of which are incorporated herein by reference.
  • Biped robots are relatively primitive compared to other types of robot movement strategies. The reason for this is due to a problem peculiar to bipedal locomotion control related to generating stable natural motion under the influence of insufficient operation and impact (for example, 1, 2, 3 of Non-Patent Document 1). See chapter).
  • Hybrid Zero Dynamics is a theoretical framework based on rigorous proof (see Non-Patent Document 2).
  • conventional HZD joint level polynomial coefficients have required time and effort during the optimization process and initialization. Then, depending on the initial value, it may be difficult to converge the optimization. For this reason, optimization convergence usually uses techniques such as first implementing a low-order polynomial for initial value grid search, etc., and then implementing a higher-order polynomial to improve the search. And improve.
  • the gait library or look-up table should be constructed to converge faster in the corresponding limit cycle.
  • the limit cycle is a closed orbit in the phase space of a dynamical system. Therefore, an alternative possibility of managing the complexity of the above HZD search has been proposed by defining a nonlinear coordinate transformation of the system state space.
  • the bipedal model is represented as an equivalent intuitive and simplified dynamic system called the Feedback Linearized Inverted Pendulum (FLIP). FLIP is not a competitor, but a complement to traditional HZD.
  • Non-Patent Document 3 the intuitive virtual constraint design for bipedal walking with 3 links is disclosed in Non-Patent Document 3, but a design method for handling 5 links or more has not been established.
  • the bipedal walking robot includes a torso, two legs, a conversion unit, and a control unit, and the two legs have a swing leg in response to walking.
  • the stances alternate, the length from the center of gravity of the two legs to the tip of the stance is r com , and the length from the center of gravity to the tip of the swing leg is rfoot .
  • the angle between the line connecting the tip of the pedestal from the center of gravity and the ground is ⁇ com , and the line connecting the tip of the pedestal from the center of gravity and the tip of the swing leg from the center of gravity are connected.
  • x is a state vector that describes two legs in joint coordinates
  • z is a state vector that describes two legs in FLIP (Feedback Linerized Inverted Pendulum) expression
  • h ⁇ (x) is [L f h 1 (x).
  • the conversion unit converts the partial linear system into a linear system using the following equation , sets ⁇ 1 to ⁇ com [rad], and sets ⁇ 2 to ⁇ ⁇ com [rad / s].
  • the control unit generates an instruction for walking motion by the two legs using the converted linear system, controls the walking motion by the two legs, and between the conversion unit and the control unit. Feedback control is performed with.
  • the control unit describes the system equation using the converted linear system as follows.
  • U in the above equation is the control vector of 4 rows and one column
  • a (x) is the following formula
  • g i ( x ) Represents the quadratic Lie derivative
  • b (x) is the following formula
  • L f 2 h 1 (x ) represents the secondary Lie derivative by f (x) with respect to h i (x)
  • q (x) is the following equation
  • p (x) is You may do so.
  • control unit has a change in the r com with respect to the locus s of the FLIP, a change in the ⁇ foot with respect to the s, and the r foot with respect to the s. And the change of the ⁇ trunk with respect to the s may be approximated at three points for control.
  • the method for controlling a bipedal walking robot is a method for controlling a bipedal walking robot including a body, two legs, a conversion unit, and a control unit.
  • the two legs alternate between the swing leg and the stance according to walking, and the length from the center of gravity of the two legs to the tip of the stance is r com , and the play from the center of gravity to the play.
  • the length of the tip of the leg is r foot
  • the angle between the line connecting the tip of the stance from the center of gravity and the ground is ⁇ com
  • the line connecting the center of gravity to the tip of the stance is a method for controlling a bipedal walking robot including a body, two legs, a conversion unit, and a control unit.
  • the two legs alternate between the swing leg and the stance according to walking, and the length from the center of gravity of the two legs to the tip of the stance is r com , and the play from the center of gravity to the play.
  • the length of the tip of the leg
  • the angle between the center of gravity and the line connecting the tip of the swing leg is ⁇ foot
  • the line connecting the center of gravity to the tip of the stance is extended from the center of gravity to the body side.
  • the angle between the torso and the torso is ⁇ trunk
  • x is a state vector that describes two legs in joint coordinates
  • z is a state vector that describes two legs in FLIP (Feedback Linerized Inverted Pendulum) representation
  • h is a state vector that describes two legs in FLIP (Feedback Linerized Inverted Pendulum) representation
  • ⁇ (x) is [L f h 1 (x) , ..., L f h 4 (x)] T (L f h 1 (x) is h i a (x) (i is an integer from 1 4)
  • f (X) represents the Lee differentiation, and T represents the translocation matrix), and the conversion unit sets the system equation as shown in the following equation.
  • the conversion unit converts the partial linear system into a linear system using the following equation , sets ⁇ 1 to ⁇ com [rad], and sets ⁇ 2 to ⁇ ⁇ com [rad / s].
  • the control unit generates an instruction for walking motion by the two legs using the converted linear system, controls the walking motion by the two legs, and between the conversion unit and the control unit. Feedback control is performed with.
  • the parameters of the bipedal walking model are selected, and the inverted pendulum model having a variable length is obtained by using feedback linearization with four variables ( ⁇ trunk , ⁇ foot ,). Performs linear control of r com , r foot).
  • (1) to (5) bipedal walking control can be considered in inverted pendulum coordinates, and adjustment in various situations becomes easy.
  • the linear control is performed approximately, the adjustment in various situations becomes easy.
  • FIG. 1 is a diagram showing a configuration example of the bipedal walking robot 1 according to the present embodiment.
  • the bipedal walking robot 1 is, for example, a humanoid robot. As shown in FIG. 1, the biped robot 1 includes a torso 11, two legs 12 (12A, 12B), a sensor 13, a drive unit 14, and a control device 15. In FIG. 1, the power supply unit, gears, and the like are omitted.
  • the bipedal walking robot 1 may include a head, two arms, and the like.
  • a control device 15, a sensor 13, a power supply unit, and the like are incorporated in the body 11.
  • the leg 12A and the leg 12B are controlled by the control device 15 so that the swing leg and the stance leg are alternately alternated according to walking.
  • the bipedal walking robot 1 is controlled so that one of the leg portion 12A and the leg portion 12B is in contact with the ground during walking.
  • the stance is the leg 12 on which the weight of the biped robot 1 is applied during walking, and the leg 12 on the side in contact with the ground.
  • the swing leg is the leg 12 on which the weight of the biped robot 1 is not applied during walking, and the leg 12 on which the biped robot 1 is not in contact with the ground. As shown in FIG.
  • each leg 12 has a first joint 121 (121A, 121B) corresponding to a human ankle, a second joint 122 (122A, 122B) corresponding to a human knee, and a hip joint connected to the torso 11.
  • 123 123A, 123B
  • the first section 124 124A, 124B
  • the second section 125 125A, 125B
  • Section 3 126 126A, 126B
  • the sensor 13 is attached to, for example, at least one of the torso 11, the nodes or joints of the legs 12.
  • the sensor 13 detects the first angle to the fifth angle. The angle detected by each sensor will be described later with reference to FIG.
  • the sensor 13 outputs angle information indicating the detected angle to the control device 15.
  • the drive unit 14 is an actuator and is driven according to the control of the control device 15.
  • the drive unit 14 is attached to each joint, for example.
  • the control device 15 includes a coordinate conversion unit 151, a control unit 152, and a drive circuit 153.
  • the control device 15 controls bipedal walking by controlling the torso 11 and the legs 12.
  • the coordinate conversion unit 151 acquires the detection value output by the sensor 13, and obtains the first angle q 1 to the fifth angle q 5 using the acquired detection value.
  • the coordinate conversion unit 151 converts the obtained first angle q 1 to fifth angle q 5 into four variables ( ⁇ trunk , ⁇ foot , r com , and r foot ) described later.
  • the coordinate conversion unit 151 converts the nonlinear system into a linear system by using the converted four variables and the input vector u output by the control unit 152.
  • the coordinate conversion unit 151 stores the formula required for the conversion.
  • the coordinate conversion unit 151 outputs the system equation of the converted linear system to the control unit 152. The method of coordinate conversion and the method of conversion to a linear system will be described later.
  • the control unit 152 calculates an indicated value that is the target trajectory of the leg unit 12 by using the system equation output by the coordinate conversion unit 151.
  • the control unit 152 outputs the calculated indicated value to the drive circuit 153.
  • the control unit 152 outputs the control vector u based on the calculated indicated value to the coordinate conversion unit 151.
  • the drive circuit 153 generates a drive signal for each drive unit 14 based on an instruction value output from the control unit 152, and outputs the generated drive signal to each drive unit 14.
  • FIG. 2 is a diagram for explaining an angle detected by the sensor 13 according to the present embodiment.
  • the leg portion 12A is a stance and the leg portion 12B is a free leg.
  • the first angle q 1 is the angle of the pedestal 12A with respect to the body 11 with the third section 126A.
  • the control device 15 obtains the first angle q 1 based on, for example, the detected value of the sensor 13 (FIG. 1) attached to the hip joint 123A (FIG. 1).
  • the second angle q 2 is the angle of the swing leg 12B with respect to the body 11 with the third section 126B.
  • the control device 15 obtains the second angle q 2 based on, for example, the detected value of the sensor 13 (FIG. 1) attached to the hip joint 123B (FIG. 1).
  • the third angle q 3 is the angle of the second section 125A with respect to the third section 126A of the pedestal 12A.
  • the control device 15 obtains the third angle q 3 based on, for example, the detected value of the sensor 13 (FIG. 1) attached to the second joint 122A (FIG. 1).
  • the fourth angle q 4 is the angle of the second section 125B with respect to the third section 126B of the swing leg 12B.
  • the control device 15 obtains the fourth angle q 4 based on, for example, the detected value of the sensor 13 (FIG. 1) attached to the second joint 122B (FIG. 1).
  • Fifth angle q 5 is the angle of inclination of the body 11 relative to the position p e.
  • the control device 15 obtains the fifth angle q 5 based on, for example, the detected value of the sensor 13 (FIG. 1) attached to the body 11.
  • the position pe is the Cartesian position of the tip of the pedestal 12A with respect to the absolute reference coordinate system O.
  • the length L1 of the fuselage 11, the length L2 of the second section 125A, the length L3 of the second section 125B, the length L4 of the third section 126A, and the length 126B of the fourth section are known and coordinate conversion.
  • the unit 151 stores.
  • FIG. 3 is a diagram for explaining an outline of a control method of the bipedal walking robot 1.
  • a SLIP Spring Loaded Inverted Pendulum
  • the SLIP model is a model that simplifies the running of humans.
  • a bipedal walking robot Boby
  • the control device 15 converts the detection value detected by the sensor 13 and describes it in the system equation, transforms the system equation into coordinates, and uses the SLIP model in the converted coordinate system to control the control value.
  • the feedback control is performed so that the bipedal walking robot follows the model by generating.
  • FIG. 4 is a diagram illustrating an outline of feedback control performed in the present embodiment and an outline of operation of the control device 15.
  • FIG. 5 is a flowchart of the processing procedure according to the present embodiment.
  • the coordinate conversion unit 151 includes a system 1511 and a conversion unit 1512.
  • the control unit 152 includes a controller 1521 and a control vector generation unit 1522.
  • the system 1511 uses the control vector u output by the control unit 152 to calculate the state vector x that describes the two legs in joint coordinates.
  • the system 1511 uses the detection values (angle q 1 to angle q 5 ) detected by the sensor 13 as four variables ( ⁇ trunk , ⁇ foot , r) using the known body 11 length and the known node length. com , rfoot ).
  • the system 1511 describes the system equation h (x) using the state vector x and four variables (step S1). The state vector, four variables, and system equations will be described later.
  • z is a state vector that describes two legs in FLIP representation. The conversion formula will be described later.
  • the controller 1521 converts the system equation into a linear system by describing the system equation in the converted coordinate system (step S3).
  • the controller 1521 outputs a vector h ... desired (denoted as h ... in FIG. 4) that specifies the desired trajectory of the FLIP system.
  • Control vector generation unit 1522 uses the h ⁇ ⁇ Desired, calculates the control vector u, and outputs the calculated control vector u to the system 1511 (step S4). After the processing, the control vector generation unit 1522 returns to the processing of step S1.
  • feedback control is performed between the system 1511 and the controller 1521 by generating a control vector using the SLIP model in the converted coordinate system.
  • Non-Patent Document 1 the equation of motion of a single support phase biped robot can be expressed in the form of a standard manipulator equation such as the following equation (1) (Non-Patent Document 1). See p.9). It is assumed that the tip of the pedestal remains stationary on the ground during a single support.
  • B is a mapping from a control signal to a state variable.
  • u is the motor torque of the drive unit 14 included in each section, and is an input vector (control vector) of 4 rows and 1 column.
  • q is a 5-by-1 vector containing q 1 , ..., Q 5.
  • q ⁇ is a vector of angular velocities corresponding to q.
  • q ... is a vector of acceleration corresponding to q.
  • G (q), C (q, q ⁇ ), and D (q) are scalar functions with q and q ⁇ as variables, respectively.
  • equation (1) is expressed as the following equation (2).
  • Equation (2) x ⁇ is a state vector and is [q T , q ⁇ T ] T. T represents the transposed matrix. Assuming that the first term on the right side of the equation (2) is f (x) and the second term is g (x) u, the equation (2) is expressed as the following equation (3) (Non-Patent Document). 1 See p.10).
  • R is a coordinate relabeling matrix that gives an angle transformation by making the previous swing leg a new stance.
  • 0 5 ⁇ 2 is a vector of 5 rows and 2 columns having 0 elements.
  • ⁇ ⁇ q ⁇ e is the following equation (5).
  • I 5 ⁇ 2 is a 5-by-2 unit vector having a diagonal element of 1.
  • D e can be acquired from the coordinate equation of the extended equation of motion (6).
  • the derivation of the formula (6) is described in Sec. Apply the procedure described in 3.4.2.
  • E 2 is the Jacobian of the Cartesian swing position with respect to q e , and ⁇ F 2 is given by Eq. (7).
  • the reaction force due to the floor (ground) (hereinafter referred to as the floor reaction force (Ground Reaction Force)) is calculated.
  • the calculation of the floor reaction force applied to the stance requires the extended equation of motion of the following equation (8) including the external force and torque ext applied to the system.
  • Equation (8) can be decomposed and expressed as in the following equation (9).
  • Equation (9) J is the contact point Jacobian matrix.
  • F ground is the floor reaction force.
  • ⁇ ext and joints are external torques applied to the joints.
  • equation (10) q ... Can be determined using equation (3).
  • the function of equation (10) is evaluated over the entire period of the simulation so as not to violate the static stance assumption.
  • FIG. 6 is a diagram for explaining four variables according to the present embodiment.
  • the CoM is the center of gravity (L centroidal ) of the two legs 12 (12A, 12B).
  • Reference numeral g11 is a line passing through the center of gravity CoM from the ground contact point of the pedestal 12A with the ground P.
  • Reference numeral g12 is a line passing through the center of gravity CoM from the ground contact point of the swing leg 12B with the ground P.
  • r com is the distance from the ground contact point of the pedestal 12A to the center of gravity CoM, and represents the length of the virtual pendulum.
  • r foot represents the distance from the center of gravity CoM to the toes of the stance 12A.
  • ⁇ foot represents the angle between the virtual pendulum and the relative position of the swing leg 12B.
  • the combination of r foot and ⁇ foot represents the position of the foot with respect to the virtual pendulum.
  • ⁇ trunk is the angle of the body 11 with respect to the virtual pendulum, and represents the inclination angle of the body with respect to the line of the symbol g11.
  • ⁇ com represents the angle between the ground P and the line of reference numeral g11. In the example of FIG. 4, ⁇ trunk and ⁇ foot are negative values, and r com , r foot, and ⁇ com are positive values.
  • the four variables of FIG. 4 ( ⁇ trunk , ⁇ foot , r com , r foot ) fully characterize the angles of all internally actuated joints.
  • the four variables do not depend on the absolute angle of the two legs described by ⁇ com. Therefore, in the present embodiment, the system 1511 converts the five detection angles (q 1 to q 5) detected by the sensor 13 into these four variables.
  • the system equation represented by these four variables can describe a variable length inverted pendulum with a specified foot arrangement.
  • the system 1511 uses the five detected detection angles (q 1 to q 5 ) to determine the center of gravity CoM.
  • the system 1511 calculates ⁇ trunk , ⁇ foot , r com , and r foot using the detected value detected by the sensor 13, the length L1 of the body 11 to be stored, and the lengths L2 to L5 of each node.
  • the coordinate transformation unit 151 calculates a r com with the length of the second section 125A L2, the length of the third section 126A L3, and the angle q 3.
  • Coordinate conversion unit 151 calculates the r foot with the length of the second section 125B L4, the length of the third section 126B L5, and the angle q 4.
  • Coordinate conversion unit 151, the length of the second section 125A L2, the length of the third section 126A L3, the angle q 3, the length of the second section 125B L4, the length of the third section 126B L5, and the angle q 4 Is used to calculate ⁇ coordinate.
  • Coordinate conversion unit 151, the length of the second section 125A L2, the length of the third section 126A L3, the angle q 3, the length of the second section 125B L4, the length of the third section 126B L5, angle q 4, And the angle q 5 is used to calculate ⁇ trunk.
  • the system 1511 describes the single-standing phase system equation as the above-mentioned equation (3).
  • the system 1511 uses the four selected variables ( ⁇ trunk , ⁇ foot , r com , r foot ) according to the procedure and notation described in Reference 1, and outputs as in the following equation (11). Define a function vector.
  • h j (x) (j is an integer from 1 to 4) is a design value and is known.
  • the columns of the matrix g (x) can be used to define the vector fields g 1 (x), g 2 (x), g 3 (x), and g 4 (x).
  • the subscript corresponds to the column g (x).
  • L g1 h 1 (x) represents that h 1 (x) is first-order Lie differentiated with respect to g 1.
  • the selected output is designed to depend only on the constituent variables (joint angles). Therefore, this condition is true and the torque of the motor does not directly affect the time derivative.
  • the reduced coupling matrix A (x) is calculated by taking the quadratic Lie derivative as in the following equation (12).
  • L g1 L f h 1 (x) represents that h 1 (x) is Lie-differentiated by f (x) and then Lie-differentiated by g 1 (x).
  • h 1 (x) is Lie-differentiated by f (x) and then Lie-differentiated by g 1 (x).
  • This state can be numerically verified and inferred during the simulation.
  • This matrix describes the effect of the control torque to the time derivative of the first case of ⁇ i ⁇ 4 L f h i (x).
  • each column of the decoupling matrix (decoupling matrix) describes whether the output time differential by the torque exerted by the motor how changes It is a vector field. Since all output functions have a relative order 2, the nonlinear coordinate transformation can be defined as in equation (13) using the output function and its first derivative.
  • the conversion unit 1512 converts the partial linear system into a linear system by converting the state vector x to z using the equation (13).
  • x is a state vector that describes the two legs in joint coordinates.
  • z is a state vector that describes two legs in FLIP representation.
  • h ⁇ (x) [L f h 1 (x), ..., L f h 4 (x)] T
  • ⁇ 1 (x) and ⁇ 2 (x) have a reversible Jacobian matrix (d ⁇ /) in ⁇ (x) at all points in the state space.
  • h ⁇ (x) is a first-order derivative of h that is known and is calculated using h, which is a design value, as will be described later.
  • ⁇ (x), ⁇ 1 (x), and ⁇ 2 (x) represent non-linear and uncontrollable components, which are unknown and are calculated by dz / dt of the equation (19) described later.
  • ⁇ 1 is set to ⁇ com [rad]. It represents the absolute angle of bipedal underactivity (and the angle of FLIP) with respect to the ground and is therefore guaranteed to have a linearly independent Jacobian.
  • the output functions are different because the selected outputs are independent with absolute biped angles.
  • ⁇ 2 is set to ⁇ ⁇ com [rad / s] and represents the rotational speed of the two legs and FLIP.
  • This is a non-linear mapping between the x and z coordinates.
  • the reversibility of d ⁇ / dx ensures that this coordinate transformation is one-to-one and the two representations are equivalent.
  • Equation (14) b (x) is the following equation (15), q (x) is the following equation (16), and p (x) is the following equation (17).
  • Equation (18) is a control equation.
  • h ⁇ desired represents the vector of 4 rows and one column, which specifies the desired trajectory of the FLIP system.
  • Equation (19) is also an equation for forward simulation of the z-coordinate system. Calculates ⁇ using d / dt [h h ⁇ ⁇ ] T in equation (19).
  • Control vector generation unit 1522 using the h ⁇ ⁇ Desired the formula (18), calculates a control vector u.
  • the parameters of the bipedal walking model are selected, the inverted pendulum model having a variable length is approximated, and the linear control of the four output functions (four variables) is approximated by using feedback linearization. To do. Then, according to the present embodiment, the bipedal walking control can be considered in the inverted pendulum coordinates, and the adjustment in various situations becomes easy.
  • the FLIP locus s can be described as a polynomial with respect to the underactuated angle normalized as in Eq. (20).
  • ⁇ com and initial are initial values of ⁇ com in a single step.
  • ⁇ com and final are the final values of ⁇ com in a single step. Therefore, s is a normalized form of ⁇ com in the range 0 to 1 and represents the phase of walking.
  • p j (s) represents a polynomial representing the nominal locus of the j-th output function of 1 ⁇ j ⁇ 4.
  • p (x) is the formula (17).
  • k j and p are position feedbacks.
  • k j and d are feedback of the derivative of the position.
  • the ⁇ trunk is controlled using feedback to keep the trunk vertical at all times.
  • the simplicity of FLIP is easily understood by designing the desired trajectories of the four outputs that have physical implications. After designing the output trajectories, the accelerations of these polynomials can also be easily obtained.
  • FIG. 7 is a diagram for explaining the polynomial shape.
  • Reference numeral g101 is a graph showing the relationship between r com and s.
  • Reference numeral g102 is a graph showing the relationship between ⁇ foot and s.
  • Reference numeral g103 is a graph showing the relationship between r foot and s.
  • Reference numeral g104 is a graph showing the relationship between ⁇ trunk and s.
  • the horizontal axis is s.
  • the vertical axis is r com .
  • the vertical axis is ⁇ foot .
  • the vertical axis in the code g103 is r foot.
  • the vertical axis is ⁇ trunk .
  • r com follows an inverted pendulum (IP), a linear inverted pendulum (LIP), a spring loaded inverted pendulum (SLIP), or a predetermined curve that can mimic any central curve.
  • IP inverted pendulum
  • LIP linear inverted pendulum
  • SLIP spring loaded inverted pendulum
  • ⁇ foot increases linearly.
  • the r foot can be approximated by a curve.
  • the r foot is acquired by a three-point fitting that includes the current foot placement, the next foot placement, and the central one to avoid obstacles.
  • the angle ⁇ trunk of the fuselage with respect to the ground, as in reference g104 can be approximated by a straight line for simplicity.
  • control is performed by approximating the change of r com with respect to s at the code g101, the change of ⁇ foot with respect to s at the code g102, the change of r foot with respect to s at the code g103, and the change of ⁇ trunk with respect to s at the code g104.
  • the coordinate transformation of the above equation (13), the control equation of the equation (18), the FLIP differential equation of the equation (19), and the like are used in the feedback control system shown in FIG. Performs bipedal control.
  • the controller 1521 controls with simpler FLIP coordinates that have been transformed, rather than the original joint coordinates.
  • Such a combination of transformation and feedback is technically also called input / output linearization.
  • FLIP coordinate transformation can provide a means of representing the state space of bipedal walking with coordinates that more conveniently represent the system. Then, in the present embodiment, the generation of the target motion can be executed in real time by using the expression of FLIP bipedal walking. And according to this embodiment, the model does not sacrifice accuracy.
  • FIG. 8 is a diagram showing the first simulation result.
  • Reference numeral g201 is a limit cycle for walking on level ground.
  • the horizontal axis is ⁇ com [rad] and the vertical axis is ⁇ ⁇ com [rad / s].
  • the symbol g202 is the average value of the cost CoT of each of the five walking types (IV).
  • the horizontal axis is the walking type and the vertical axis is the cost CoT.
  • the bumpy height is less than 0.03 m and the slope of the hilly slope is 0.1745 [rad].
  • the cost CoT is a value obtained by dividing the value obtained by integrating the angular velocities of each joint by the value obtained by multiplying the weight of the biped robot 1 by the walking distance.
  • FIG. 9 shows the result of tracking the output of flat ground walking (hereinafter referred to as flat ground walking).
  • FIG. 9 is a diagram showing a simulation result of a tracking plot of output in walking on level ground.
  • Reference numeral g301 is a graph showing the relationship between r com [m] and s.
  • the reference numeral g3011 indicates the change in r com of the simulation result, and the reference numeral g3012 indicates the expected value of r com.
  • Reference numeral g302 is a graph showing the relationship between ⁇ foot [rad] and s.
  • the reference numeral g3021 indicates the change in ⁇ foot of the simulation result, and the reference numeral g3022 indicates the expected value of ⁇ foot.
  • Reference numeral g303 is a graph showing the relationship between r foot [m] and s.
  • Reference numeral g3031 indicates a change in r foot of the simulation result, and reference numeral g3032 indicates an expected value of r foot.
  • Reference numeral g304 is a graph showing the relationship between ⁇ trunk [rad] and s.
  • the reference numeral g3041 indicates a change in ⁇ trunk of the simulation result, and the reference numeral g3012 indicates an expected value of ⁇ trunk.
  • the horizontal axis is s.
  • the vertical axis is r com .
  • the vertical axis is ⁇ foot .
  • the vertical axis in the code g303 is r foot.
  • the vertical axis is ⁇ trunk .
  • the simulation results and expected values match or almost match, as in the symbols g3011 and g3012, g3021 and g3022, g3031 and g3032, and g3042 and g3041.
  • r com is related to the straightness of the leg during walking and the length of the virtual pendulum in the FLIP representation of bipedal walking.
  • the walking speed can be adjusted by increasing or decreasing this value at the midpoint of walking.
  • ⁇ foot is related to foot placement, step distance, and walking speed.
  • the final value of rfoot during a step in gait control is a factor in gait speed and stability, which affects how much the model has moved "forward" at the end of each step, from one step. This is because it represents the amount of momentum to move to the next step.
  • the advantage of using FLIP coordinates is that the intuitive interpretation of FLIP variables allows the required motion to be generated in real time. Then, according to the present embodiment, by interpolating the start point, middle point and end point of the orbital polynomial so as to satisfy the continuity from the previous step so as to compensate for any large change in the basic height. , The desired dynamic characteristics (walking speed, ground reaction force, etc.) can be achieved in the current step. All required motor torque (saturation 200 Nm) and foot contact friction (maximum value 0.6) are inspected in the simulation process to meet practical values.
  • ⁇ Initial value or model parameter> The change of the output trajectory of the FLIP coordinates by the controller 1521 depends on the initial value of the FLIP space. As shown in FIG. 9, for example, if the initial value of r com is 0.815 [m], the controller 1521 has three points (0, 0.815) (reference numeral g3013), (0.5, 0.815-deviation) (code G3014), determines the desired trajectory r com in (1,0,815) (code G3015). (Deviation is a constant.
  • FIG. 10 is a diagram showing a comparative example of different initial values of r com.
  • reference numeral g401 is a graph showing a comparison of limit cycles of each of different initial values of r com.
  • the horizontal axis is ⁇ com [rad]
  • the vertical axis is ⁇ ... com [rad / s].
  • Reference numerals g402 and g403 are graphs showing a comparison of one-cycle data for control variables and non-control variables.
  • the horizontal axis is time [s] and the vertical axis is the expected value of r com [m].
  • reference numeral g403 the horizontal axis is time [s] and the vertical axis is ⁇ com [rad / s].
  • the initial value is changed, but the simulation results, such as the code 4011 and G4012, the initial value is made form adapted to a stable limit cycle with respect to two different r com ..
  • the controller 1521 can generate a new target trajectory and maintain walking.
  • the initial value r com is changed to 0.735 [m]
  • the length of the pendulum is contracted, so that the person is in a state of crouching and walking.
  • FIG. 11 is a diagram showing other parameters during gait control.
  • Reference numeral g501 is a graph showing the transition between ⁇ com and ⁇ ... com , which is ⁇ com [rad], and the vertical axis is ⁇ ... com [rad].
  • Reference numerals g502 to g504 are control signals for the steps.
  • the horizontal axis is a step.
  • the vertical axis is vGRF (floor reaction force vertical component; vertical ground reaction force) [N].
  • the vertical axis is torque [Nm].
  • Reference numeral g5031 is a waist torque u1 when standing.
  • Reference numeral g5032 is the torque u2 of the knee joint when standing.
  • Reference numeral g5041 is a waist torque u3 when walking.
  • Reference numeral g5042 is the torque u4 of the knee joint when walking.
  • the controller 1521 does not need to change the configuration of the FLIP model and the controller (see FIG. 4), for example, shortening the rcom.
  • a new walking model can be created directly by simply adjusting the initial values.
  • the vertical component of the floor reaction force in the steady state is different from that described in Non-Patent Document 1 (p.185).
  • FIG. 12 is a diagram showing a limit cycle and a control signal when the walking mode is changed.
  • Reference numeral g601 is a graph showing a limit cycle in the transition from backward walking to forward walking, which is ⁇ com [rad], and the vertical axis is ⁇ ... com [rad].
  • Reference numerals g602 and g603 are control signals for the steps.
  • the horizontal axis is a step.
  • the vertical axis is the expected value of r com [m].
  • the vertical axis is the expected value of r foot [m].
  • the symbols g6021 and g6031 represent an uphill walk.
  • the symbols g6022 and g6032 represent walking on level ground.
  • the symbols g6023 and g6033 represent walking downhill.
  • the figure shown by reference numeral g602 shows that three important steps are shown during walking from up to down.
  • the first step is a standard uphill walk that requires control of r com and r foot.
  • the second step is the final step of the above, the deviation from the midpoint and the initial point of the r foot not changed r com is changed from -0.2 to -0.15.
  • the third step when the model first walked down, the r com changed to a lower height and the midpoint deviation of the r foot was changed to -0.1.
  • the figure shown by the symbol g603 shows that the model can dynamically shift from the rear to the front by simply changing the swing direction of the ⁇ foot and changing the sign of the initial angular velocity of the inverted pendulum (for example, +1 to -1). Is shown.
  • VSA variable rigidity joints
  • FIG. 13 is a diagram showing the results of simulating walking on a flat ground by a bipedal robot with a variable rigidity joint.
  • Reference numeral g701 is a graph showing the relationship between the rigidity [Nm / rad] of the knee joint of the stance with respect to s (phase of walking).
  • Reference numeral g702 is a graph showing the relationship between s and r com [m].
  • Reference numeral g703 is a graph showing the relationship of the rigidity [Nm / rad] of the knee joint of the swing leg with respect to s.
  • Reference numeral g704 is a graph showing the relationship between s and r foot [m].
  • the horizontal axis is s.
  • the vertical axis represents the rigidity [Nm / rad] of the knee joint of the stance.
  • the vertical axis is r com [m].
  • the vertical axis represents the rigidity [Nm / rad] of the knee joint of the swing leg.
  • the vertical axis is r foot [m]. From FIG. 13, it can be seen that for vibration-free gait, a more complex knee stiffness profile needs to be designed.
  • FLIP allows intuitive adjustment of complex gait models is that FLIP transforms complex joint space into simple FLIP space, and the desired trajectory of the output is monotonous or convex, so search. This is because the process is greatly simplified. Optimization may be performed in FLIP.
  • the simulation confirmed the efficiency, adaptability, and ease of controller design realized by the model.
  • the FLIP model is used for the walking control of the bipedal walking robot 1. Then, in the present embodiment, a physical model of plane 5-link bipedal walking is introduced, and FLIP transformation is analytically derived and coordinate transformation is performed by using the method of nonlinear control theory. FLIP enables intuitive tuning and can also be learned by neural networks.
  • feedback linearization is used to enable linear control of four variables and approximately describe a variable length inverted pendulum.
  • the bipedal walking control can be completely executed by the pendulum coordinates, it is easy to adjust according to various tasks, the transportation cost (low cost of transportation) is kept low, and it is superior to the passive walking. It is versatile.
  • bipedal walking control can be considered in inverted pendulum coordinates, and adjustment in various situations becomes easier than before.
  • control device 15 By recording a program for realizing a part or all of the functions of the control device 15 in the present invention on a computer-readable recording medium, and having the computer system read and execute the program recorded on the recording medium. A part or all of the processing performed by the control device 15 may be performed.
  • the term "computer system” as used herein includes hardware such as an OS and peripheral devices.
  • the "computer system” shall also include a WWW system provided with a homepage providing environment (or display environment).
  • the "computer-readable recording medium” refers to a storage device such as a flexible disk, a magneto-optical disk, a portable medium such as a ROM or a CD-ROM, or a hard disk built in a computer system.
  • a "computer-readable recording medium” is a volatile memory (RAM) inside a computer system that serves as a server or client when a program is transmitted via a network such as the Internet or a communication line such as a telephone line.
  • RAM volatile memory
  • the above program may be transmitted from a computer system in which this program is stored in a storage device or the like to another computer system via a transmission medium or by a transmission wave in the transmission medium.
  • the "transmission medium” for transmitting a program refers to a medium having a function of transmitting information, such as a network (communication network) such as the Internet or a communication line (communication line) such as a telephone line.
  • the above program may be for realizing a part of the above-mentioned functions. Further, it may be a so-called difference file (difference program) that can realize the above-mentioned function in combination with a program already recorded in the computer system.

Landscapes

  • Engineering & Computer Science (AREA)
  • Robotics (AREA)
  • Mechanical Engineering (AREA)
  • Manipulator (AREA)
  • Feedback Control In General (AREA)

Abstract

Un robot marcheur bipède comprend : une partie corps ; deux parties jambe ; une unité de conversion ; et une unité de commande. Les deux parties jambe fonctionnent en alternance comme jambe livre et jambe au sol à mesure que le robot marche. L'unité de conversion définit une équation de système en utilisant rcom, rpied, thêtapied et thêtatronc. L'unité de conversion convertit un système linéaire partiel en un système linéaire. L'unité de commande génère, à l'aide du système linéaire obtenu par conversion, une instruction pour un mouvement de marche avec les deux parties jambe, commande le mouvement de marche avec les deux parties jambe, et exécute une commande de retour entre l'unité de conversion et l'unité de commande.
PCT/JP2020/039605 2019-11-11 2020-10-21 Robot marcheur bipède et procédé de commande de robot marcheur bipède WO2021095471A1 (fr)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
JP2019204046A JP7133533B2 (ja) 2019-11-11 2019-11-11 二足歩行ロボット、および二足歩行ロボットの制御方法
JP2019-204046 2019-11-11

Publications (1)

Publication Number Publication Date
WO2021095471A1 true WO2021095471A1 (fr) 2021-05-20

Family

ID=75899928

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/JP2020/039605 WO2021095471A1 (fr) 2019-11-11 2020-10-21 Robot marcheur bipède et procédé de commande de robot marcheur bipède

Country Status (2)

Country Link
JP (1) JP7133533B2 (fr)
WO (1) WO2021095471A1 (fr)

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114527663B (zh) * 2022-02-24 2023-03-14 北京理工大学 基于变参数速度映射和二次规划的双足机器人落脚点规划方法

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2005199383A (ja) * 2004-01-15 2005-07-28 Sony Corp 動的制御装置および動的制御装置を用いた２足歩行移動体
JP2011025339A (ja) * 2009-07-23 2011-02-10 Honda Motor Co Ltd 移動体の制御装置
JP2018185747A (ja) * 2017-04-27 2018-11-22 国立大学法人京都大学 非線形システムの制御方法、二足歩行ロボットの制御装置、二足歩行ロボットの制御方法及びそのプログラム

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2005199383A (ja) * 2004-01-15 2005-07-28 Sony Corp 動的制御装置および動的制御装置を用いた２足歩行移動体
JP2011025339A (ja) * 2009-07-23 2011-02-10 Honda Motor Co Ltd 移動体の制御装置
JP2018185747A (ja) * 2017-04-27 2018-11-22 国立大学法人京都大学 非線形システムの制御方法、二足歩行ロボットの制御装置、二足歩行ロボットの制御方法及びそのプログラム

Also Published As

Publication number Publication date
JP2021074828A (ja) 2021-05-20
JP7133533B2 (ja) 2022-09-08

Similar Documents

Publication Publication Date Title
Al-Shuka et al. Multi-level control of zero-moment point-based humanoid biped robots: a review
Lee et al. Development, analysis, and control of series elastic actuator-driven robot leg
Kenwright Controlled biped balanced locomotion and climbing
Yi et al. Learning full body push recovery control for small humanoid robots
US20240000648A1 (en) Method for moving an exoskeleton
Azimi et al. Model-based adaptive control of transfemoral prostheses: Theory, simulation, and experiments
Shafiee et al. Online dcm trajectory generation for push recovery of torque-controlled humanoid robots
Reher et al. Inverse dynamics control of compliant hybrid zero dynamic walking
Yamakita et al. Extended passive velocity field control with variable velocity fields for a kneed biped
WO2021095471A1 (fr) Robot marcheur bipède et procédé de commande de robot marcheur bipède
Verrelst et al. Control architecture for the pneumatically actuated dynamic walking biped “Lucy”
Liu et al. Workspace trajectory generation method for humanoid adaptive walking with dynamic motion primitives
Lim et al. Control design to achieve dynamic walking on a bipedal robot with compliance
Xie et al. Gait optimization and energy-based stability for biped locomotion using large-scale programming
Ahmed et al. Variable admittance control for climbing stairs in human-powered exoskeleton systems
Ahmed et al. Modified admittance control for maneuverable human-powered augmentation lower exoskeleton systems
Ratre et al. Offline Trajectory Generation for Bipedal Robot using Linear Inverted Pendulum Model
Guihard et al. Simulation of a dynamic vertical jump
Pournazhdi et al. Dynamic modeling and sliding mode control for fast walking of seven-link biped robot
Araffa et al. Implementation and simulation a model predictive control for motion generation of biped robot
Ayari et al. The generation of a stable walking trajectory of a biped robot based on the COG based-gait pattern and ZMP constraint
Kee et al. Biologically inspired joint control for a humanoid robot
Roos Design of a direct collocation model predictive control framework for legged locomotion based on whole-body dynamics and explicit contact phases
Sari et al. Implementation and integration of fuzzy algorithms for descending stair of KMEI humanoid robot
Manchester et al. Stable dynamic walking over rough terrain: Theory and experiment

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 20887961

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 20887961

Country of ref document: EP

Kind code of ref document: A1