A humanoid hip joint parallel drive robot lower limb and its control method
By using a three-axis coaxial spherical parallel drive hip joint mechanism and reinforcement learning control, the problems of concentrated driving torque, reduced dynamic response and insufficient structural compactness in traditional serial hip joint structures have been solved, achieving high power density and high precision hip joint motion control.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- SHENYANG INST OF AUTOMATION - CHINESE ACAD OF SCI
- Filing Date
- 2026-03-02
- Publication Date
- 2026-05-05
AI Technical Summary
Traditional tandem hip joint layouts have highly concentrated driving torques during primary load-bearing movements such as flexion and extension, resulting in large size and high mass of motors and reduction mechanisms, and limited power density. The abduction, extension, and rotational degrees of freedom need to support the mass and inertia of the upstream joints, which reduces dynamic response capability. There is inertial coupling and error accumulation between the degrees of freedom, which affects the overall stiffness of the joint, control accuracy and dynamic stability. The joint structure has a long dimension along the limb axis, which is not conducive to compact design and biomimetic shape.
A spherical parallel drive hip joint mechanism with three coaxial axes is adopted. Multiple drive units act in parallel on the same spherical output component to achieve coordinated drive of flexion, extension, abduction, and rotation. Combined with linear drive and linkage mechanism, it forms a parallel hip joint system, and reinforcement learning control is introduced to compensate for model error and external disturbance.
It improves drive utilization, enhances dynamic performance, achieves structural compactness, improves joint motion precision and stability, and strengthens robustness and dynamic balance in complex terrain.
Smart Images

Figure CN121757299B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of robot joints and bionic motion mechanisms, and in particular to a humanoid hip joint parallel drive robot lower limb and its control method. Background Technology
[0002] The lower limbs of humanoid robots can mimic human walking and standing, achieving stability, flexibility and efficiency in movement. Traditional humanoid robot lower limbs mostly adopt a series configuration, which is more intuitive and relatively simple to design and control. According to the different arrangement order of the hip joint based on the rotation axis, there are three different arrangement methods: FAR (flexion-extension, abduction-introversion, rotation), RAF (rotation, abduction-introversion, flexion-extension), and ARF (abduction-introversion, flexion-extension, rotation). The advantages and disadvantages of each arrangement method are closely related to their axis sequence and motion mode. (1) FAR (flexion-extension, abduction-introversion, rotation) controls flexion-extension (F) first, then abduction-introversion (A), and finally rotation (R). Flexion-extension is prioritized, which conforms to the natural human movement; abduction-introversion is moderate; rotation is appropriately delayed, without interfering with other movements. However, the rotation response is slightly inferior, and the coupling of multiple degrees of freedom is complex. (2) RAF (rotation, abduction-introversion, flexion-extension) controls rotation (R) first, then abduction-introversion (A), and finally flexion-extension (F). Rotation is prioritized, which helps dynamic stability and gait adjustment; the abduction-introversion degree of freedom is appropriate. However, the delayed flexion and extension degrees of freedom may lead to a delay in flexion and extension movements; the degree of freedom coupling is relatively strong. (3) ARF (abduction, flexion and extension, rotation) controls abduction (A) first, then flexion and extension (F), and finally rotation (R). Prioritizing abduction helps with balance adjustment; the flexion and extension degrees of freedom are flexible; rotation is precise and independent. However, prioritizing abduction may affect gait stability; the rotation response is not flexible enough.
[0003] In traditional tandem hip joint designs, the three rotational degrees of freedom are achieved sequentially via a series drive chain. While this structure is relatively simple in implementation, it suffers from the following shortcomings in practical applications:
[0004] (1) In the main load-bearing actions such as bending and stretching, the driving torque is highly concentrated in a single driver, resulting in large size and high mass of the motor and reduction mechanism, and limited power density;
[0005] (2) The abduction, inward and outward and rotational degrees of freedom need to bear the mass and inertia of the upstream joint, which leads to a decrease in dynamic response capability and is not conducive to rapid gait adjustment and balance control;
[0006] (3) Significant inertial coupling and error accumulation exist between the degrees of freedom in the series structure, affecting the overall stiffness, control accuracy and dynamic stability of the joint;
[0007] (4) The joint structure has a long dimension along the limb axis, which is not conducive to compact design and biomimetic shape realization.
[0008] Therefore, there is an urgent need for a new type of hip joint mechanism that can meet the multi-degree-of-freedom motion requirements of the hip joint while improving drive utilization, enhancing dynamic performance, and achieving a compact structure. Summary of the Invention
[0009] To address the shortcomings of existing technologies, this invention provides a humanoid hip joint parallel drive robot lower limb and its control method. This invention proposes a humanoid hip joint parallel drive, providing a spherical parallel drive hip joint mechanism with a three-axis coaxial arrangement. Multiple drive units act in parallel on the same spherical output component to achieve coordinated drive of flexion, extension, abduction, and rotation, thereby overcoming the shortcomings of existing serial hip joint structures in terms of torque distribution, dynamic performance, and structural compactness.
[0010] On one hand, the present invention provides a humanoid hip joint parallel drive robot lower limb, including a parallel hip joint system, a thigh-knee joint system and a lower leg-ankle joint system.
[0011] The parallel hip joint system includes a hip joint motor base, a rotary joint module, an inner output link, a middle output link, an outer output link, an inner connecting link, a middle connecting link, an outer connecting link, a hip joint output flange, a bearing, a bearing end cover, and a bearing central shaft.
[0012] The rotary joint module has three sets, which are fixed to the hip joint motor base. The output shafts of the three sets of rotary joint modules are respectively connected to one end of the inner output link, the middle output link and the outer output link. The other ends of the inner output link, the middle output link and the outer output link are respectively connected to one end of the inner connecting link, the middle connecting link and the outer connecting link through bearings. The other ends of the inner connecting link, the middle connecting link and the outer connecting link are connected to the hip joint output flange through bearings, forming a parallel mechanism of three chains.
[0013] The three sets of rotary joint modules are the inner rotary joint module, the middle rotary joint module and the outer rotary joint module, each of which includes an independent drive motor and a reduction mechanism; the inner output link, the middle output link and the outer output link are respectively connected to the inner rotary joint module, the middle rotary joint module and the outer rotary joint module.
[0014] The bearing, bearing end cap, and bearing central shaft together form a rotational hinge for the drive link, which is connected between the outer connecting link and the hip joint output flange to achieve hinge connection.
[0015] The thigh-knee joint system includes a thigh structural frame, a thigh linear joint module, a thigh link, a knee link, and a lower leg structural frame.
[0016] The thigh structural frame is a hollow structure, connected to the hip joint output flange. A linear drive module is installed inside the thigh structural frame and transmits power to the lower leg structural frame. The linear drive module achieves precise linear motion servo through motor motion and feedback control.
[0017] The thigh linear joint module is placed inside the thigh structural frame, and its two ends are respectively connected to the root of the thigh structural frame and the pivot of the root of the knee joint link through hinges.
[0018] The output shaft of the thigh linear joint module is connected to the thigh link, the thigh link is hinged to the knee link, and the knee link is hinged to the lower leg structural frame.
[0019] The lower leg-ankle joint system includes a medial lower leg linear joint module, a lateral lower leg linear joint module, an ankle joint base, an ankle joint universal joint, and a foot.
[0020] The inner lower leg linear joint module and the outer lower leg linear joint module are respectively arranged on both sides of the lower leg structural frame; the two ends of the inner lower leg linear joint module and the outer lower leg linear joint module are respectively connected to the axis of the lower leg structural frame and the ankle joint base by hinge.
[0021] The ankle joint base receives signals from the linear drive module and connects the lower leg structural frame to one end of the ankle joint universal joint; the other end of the ankle joint universal joint is connected to the foot.
[0022] The inner rotary joint module, middle rotary joint module, outer rotary joint module, thigh linear joint module, inner calf linear joint module, and outer calf linear joint module integrate motors, reducers, drivers, encoders, and force sensors inside the joints; the hip joint motor base is fixedly mounted with an inertial measurement unit (IMU).
[0023] On the other hand, the present invention also provides a control method for the lower limb of a humanoid hip joint parallel drive robot, which is implemented through the aforementioned humanoid hip joint parallel drive robot lower limb, specifically including:
[0024] Construct a kinematic model;
[0025] The hip joint parallel system, based on the characteristics of the human leg hip joint, adopts a coaxial drive layout, thus solving the problem of limited pitch angle. All rotational joint axes intersect at a single point, defined as the spherical rotation center, i.e., the origin of the mechanism. The angle between the inner connecting rod and the middle connecting rod is... One end is connected to three sets of rotary joint modules respectively; the vector of the connected joint axis is defined as , This indicates three branches, with the other end hinged to the middle connecting rod, and the curvature is... , The angle is represented by the half-cone angle of the hip joint output flange, and the connected joint axis vector is defined as... Together, they form a branch; the branches are spaced 120° counterclockwise, and the branches are connected to the hip joint output flange via a revolute joint; the axis vector of the hip joint output flange shaft is defined as... The included angle between different rotational joint axes is , Define the origin of the world coordinate system. Construct a three-dimensional orthogonal right-handed coordinate system at the center point of the hip joint output flange plane. The shaft is perpendicular to the hip joint output flange. The axis is the negative direction of the joint axis vector. The axis is determined by the right-hand rule; for branches The corresponding mechanism parameters include: the structural angle of the drive linkage. , representing a unit vector and The included angle; the structural angle of the follower link. , representing a unit vector and The included angle; the semi-cone angle of the static platform , representing a unit vector and The included angle; projecting the static platform onto On the plane formed by the axes, the distribution angles of each branch after projection are represented by parameters. It means that Axis to Unit Vector Projection lines in the world coordinate system The angle between the two sides is defined as the positive direction of rotation, with the counterclockwise direction being the positive direction.
[0026] Using the aforementioned kinematic model, kinematic solutions are calculated, specifically including inverse kinematics and forward kinematics:
[0027] The inverse kinematics solution specifically involves: based on the given hip joint output flange posture angle... , , Solve the output angle of the rotary joint module The inverse kinematics can be obtained from the closed-loop condition of the mechanism's branches;
[0028] Unit vector of each branch Represented as:
[0029] ;
[0030] Unit vector , indicating the axis where the intermediate rotational joint is located. The angle is output by the rotary joint module. have to:
[0031] ;
[0032] In the formula, Indicates about the axis Rotation angle The rotation transformation matrix, unit vector The axis containing the rotary joint connected to the hip joint output flange is indicated using Euler angles. Rotation transformation matrix of the sequence To describe the hip joint output flange orientation:
[0033] ;
[0034] ;
[0035] This represents the unit vector of the axis of the rotational joint connected to the hip joint output flange when the hip joint output flange is in its initial pose. ;
[0036] The projection of the hip joint output flange (1-11) is denoted as ,in for:
[0037] ;
[0038] Rotation axis unit vector , The constraint of a closed-loop mechanism must be satisfied:
[0039] , ;
[0040] in According to the Law of Cosines for the sphere: ;
[0041] The output angle solution of the three sets of rotary joint modules is expressed as follows:
[0042] , ;
[0043] The kinematic solution specifically refers to: based on the angles output by the three sets of rotational joint modules. Solve for the three-axis attitude angles of the hip joint output flange. The maximum number of forward kinematic solutions for a three-degree-of-freedom spherical parallel mechanism is eight, thus allowing us to derive the forward kinematic solutions.
[0044] First, establish a unit vector representing the rotation axis of each branch. in, The expression form corresponds to the inverse kinematics solution, while the expression for the unit vector is the same. When considering the overall integrity of the hip joint output flange, the coordinate transformation relationship should be adjusted accordingly. Represented as:
[0045] ;
[0046] Introducing unknowns , defined as the rotation angle of the middle rotation axis of the first branch, simplifies by substituting into the rotation matrix:
[0047] ;
[0048] Where vectors are used This represents a known parameter, specifically as follows:
[0049] , , ;
[0050] Therefore, the unit vector of the output axis of the second branch Represented as: ;
[0051] Substituting the rotation matrix and simplifying, we get: ;
[0052] Where vectors are used This represents a known parameter, specifically as follows:
[0053] , , ;
[0054] because They are all fixed to the hip joint output flange, so Decomposed into: ;
[0055] in For constant scalar, according to Calculation, that is: , , ;
[0056] Based on the constraints of institutional closed chains, establish and The two constraint equations in:
[0057] , ;
[0058] This leads to the information about the unknown quantity. The system of equations:
[0059] ;
[0060] in , , , , , ;
[0061] Next, we will discuss the unknown quantities. The system of equations can be transformed into the following form:
[0062] ;
[0063] according to , and then , ,in After simplification, we get The octet equation in one variable is: ;
[0064] Solve the octet equation in one variable to obtain the analytical expression for y, then: ;
[0065] Therefore, obtain ;
[0066] according to Inverse Euler angle solution, if: ;
[0067] but This is the correct kinematic solution.
[0068] The aforementioned humanoid hip joint parallel-driven robot lower limb further includes a control system for controlling the humanoid robot lower limb system; the control system specifically includes:
[0069] (1) Gait and task generation module: Based on the robot's walking task, generate the expected trajectory of the foot in space and the reference motion of the robot's center of mass;
[0070] (2) Series-parallel hybrid lower limb kinematics solution module: A hierarchical kinematics solution method was established: First, based on the desired pose of the foot and the position of the pelvis, the geometric configuration of the entire lower limb in space was calculated. Then, the geometric configuration was decomposed into three parts: hip posture, knee flexion, and ankle posture. For the hip and ankle joints, the inverse kinematics model was used to map the required spatial posture into the displacement or angle of each driving branch. For the knee joint, the corresponding joint angle was obtained directly through the geometric relationship between the knee and ankle joints.
[0071] (3) Cooperative control module: obtains the target position, velocity or torque of each drive actuator, generates executable motion commands based on the geometric model and structural constraints of the lower limb, and ensures that the overall posture of the lower limb meets the planned gait and foot trajectory requirements;
[0072] (4) Reinforcement learning control module: Taking the real-time state of the robot's lower limbs as input, it outputs compensation quantities for the control quantities of each joint through reinforcement learning, correcting the errors caused by the kinematic model and dynamic model;
[0073] The reinforcement learning formula is as follows: ;
[0074] Where u represents the system control input, including the control inputs for the three rotary motors of the hip joint and the three linear motors of the knee and ankle joints. For stabilizing control items, To enhance learning output, Adaptive weights related to structural risk;
[0075] The kinematic model is responsible for providing basic motion commands that conform to the mechanism constraints. Based on this, compensation is made for execution errors, ground changes and dynamic uncertainties, and the final control commands are formed by superimposing them.
[0076] (5) Low-level execution and closed-loop feedback module: The final control command is sent to the inner rotary joint module, the middle rotary joint module, the outer rotary joint module, the thigh linear joint module, the inner lower leg linear joint module, and the outer lower leg linear joint module. The module outputs torque or displacement according to the control command, so that the robot's lower limbs produce corresponding movements. At the same time, sensor information is continuously collected and the feedback information is sent to the reinforcement learning control module and the serial-parallel hybrid lower limb kinematics calculation module, thereby forming a closed-loop control system, so that the lower limb movement can be corrected in real time and run stably.
[0077] The beneficial effects of adopting the above technical solution are as follows:
[0078] This invention provides a lower limb of a humanoid hip joint parallel drive robot and its control method. The parallel drive structure enables multiple drive units to perform parallel actions on the same output component, effectively reducing the peak torque requirement of a single driver and improving the system power density.
[0079] For flexion and extension movements that require the maximum output torque and power, the coaxial parallel drive structure can utilize three motors simultaneously to achieve maximum power density output.
[0080] Each degree of freedom no longer needs to bear the entire mass and inertia of the upstream joint, significantly improving the dynamic response performance of abduction, inward and outward and rotational movements;
[0081] By using a triaxial coaxial arrangement, a compact structural design is achieved, making the joint rotation center closer to the ideal anatomical center and improving the biomimetic effect.
[0082] The inherent high stiffness of parallel mechanisms reduces error transmission and structural compliance, thereby improving the accuracy and stability of joint movements.
[0083] By combining model-based kinematics calculation with reinforcement learning control, this invention can adaptively compensate for model errors, uneven ground, and external disturbances while ensuring the basic motion accuracy of the lower limbs, thereby significantly improving the robustness and dynamic balance of the humanoid robot's lower limbs during walking and standing.
[0084] By employing reinforcement learning to optimize joint control quantities online or offline, lower limb motion control can continuously adjust itself according to the environment and motion state, reducing energy consumption and improving motion smoothness while ensuring stability. This approach is suitable for the motion control needs of humanoid robots in complex terrains. Attached Figure Description
[0085] Figure 1 A schematic diagram of the overall structure of the lower limb of a humanoid hip joint parallel drive robot according to an embodiment of the present invention;
[0086] Figure 2 Schematic diagram of the parallel hip joint system of this invention;
[0087] Among them, 1-1 hip joint motor base, 1-2 inner rotary joint module, 1-3 middle rotary joint module, 1-4 outer rotary joint module, 1-5 inner output link, 1-6 middle output link, 1-7 outer output link, 1-8 inner connecting link, 1-9 middle connecting link, 1-10 outer connecting link, 1-11 hip joint output flange, 1-12 bearing, 1-13 bearing end cover, 1-14 bearing central shaft;
[0088] Figure 3 Schematic diagram of the thigh-knee joint system structure according to an embodiment of the present invention;
[0089] Among them, 2-1 thigh structural frame, 2-2 thigh linear joint module, 2-3 thigh connecting rod, 2-4 knee joint connecting rod, and 2-5 lower leg structural frame;
[0090] Figure 4 Schematic diagram of the lower leg-ankle joint system structure according to an embodiment of the present invention;
[0091] Among them, 3-1 medial lower leg linear joint module, 3-2 lateral lower leg linear joint module, 3-3 ankle joint base, 3-4 ankle joint universal joint, and 3-5 foot;
[0092] in Figure 4 (a) - Side view of the lower leg-ankle joint system structure. Figure 4 (b) - Rear view of the lower leg-ankle joint system;
[0093] Figure 5 The flexion and extension movements in this embodiment of the invention;
[0094] in Figure 5 (a) - High point of flexion and extension movements, Figure 5 (b) - Position after flexion and extension movements;
[0095] Figure 6 The embodiments of the present invention include rotation and abduction movements;
[0096] in Figure 6 (a) - High point of flexion and extension movements, Figure 6 (b) - Position after flexion and extension movements;
[0097] Figure 7 Schematic diagram of the kinematic model of an embodiment of the present invention;
[0098] Figure 8 A structural block diagram of a control method for the lower limbs of a humanoid hip joint parallel drive robot according to an embodiment of the present invention. Detailed Implementation
[0099] Example 1:
[0100] On one hand, the present invention provides a humanoid hip joint parallel-driven robot lower limb, such as... Figure 1 As shown, it includes a parallel hip joint system, a thigh-knee joint system, and a lower leg-ankle joint system. By introducing a three-drive parallel spherical mechanism at the hip joint, coordinated drive control of the three degrees of freedom—flexion / extension, abduction / extension, and rotation—is achieved. Linear drives and linkage mechanisms are combined at the knee and ankle joints to achieve high load-bearing capacity and good motion performance, thus forming a high-power-density, high-dynamic-performance, and highly biomimetic humanoid robot lower limb system.
[0101] The parallel hip joint system, such as Figure 2 As shown, the structure adopts a three-axis coaxial arrangement of spherical parallel structure, including a hip joint motor base 1-1, three sets of rotary joint modules, inner output link 1-5, middle output link 1-6 and outer output link 1-7, inner connecting link 1-8, middle connecting link 1-9, outer connecting link 1-10, hip joint output flange 1-11, bearing 1-12, bearing end cover 1-13 and bearing central shaft 1-14;
[0102] Three sets of rotary joint modules are fixed to the hip joint motor base 1-1. The output shafts of the three sets of rotary joint modules are respectively connected to one end of the inner output link 1-5, the middle output link 1-6, and the outer output link 1-7. The other ends of the inner output link 1-5, the middle output link 1-6, and the outer output link 1-7 are respectively connected to one end of the inner connecting link 1-8, the middle connecting link 1-9, and the outer connecting link 1-10 through bearings. The other ends of the inner connecting link 1-8, the middle connecting link 1-9, and the outer connecting link 1-10 are connected to the hip joint output flange 1-11 through bearings. Each drive chain acts in parallel to the hip joint output flange through a linkage mechanism, forming a complete three-chain parallel mechanism to achieve coordinated control of the hip joint posture.
[0103] The three sets of rotary joint modules include an inner rotary joint module 1-2, a middle rotary joint module 1-3, and an outer rotary joint module 1-4, each containing an independent drive motor and a reduction mechanism, each serving as a parallel drive branch; the inner output link 1-5, the middle output link 1-6, and the outer output link 1-7 are respectively connected to the inner rotary joint module 1-2, the middle rotary joint module 1-3, and the outer rotary joint module 1-4, and are used to convert the rotary drive into a spatial constraint effect on the output flange.
[0104] The bearing 1-12, bearing end cover 1-13, and bearing central shaft 1-14 together form a rotational hinge for the drive link, which is connected between the outer connecting link 1-10 and the hip joint output flange 1-11 to achieve hinge connection, thereby ensuring high-precision alignment and reliable rotation of the link axis.
[0105] The three-chain parallel mechanism works together through geometric constraints on the hip joint output flange 1-11 to achieve coordinated driving of the three degrees of freedom of the hip joint: flexion / extension, abduction / introduction, and rotation. During high-load movements such as flexion / extension, the three sets of drives can output torque simultaneously, thereby significantly improving the load-bearing capacity and dynamic performance of the hip joint.
[0106] The thigh-knee joint system, such as Figure 3 As shown, it includes a thigh structural frame 2-1, a thigh linear joint module 2-2, a thigh link 2-3, a knee joint link 2-4, and a lower leg structural frame 2-5.
[0107] The thigh structure frame 2-1 is a hollow structure, connected to the hip joint output flange 1-11. A linear drive module is installed inside the thigh structure frame 2-1 and transmits power to the lower leg structure frame 2-5. The linear drive module is a module that realizes high-precision, high-thrust linear motion, and achieves precise linear motion through motor motion and feedback control.
[0108] The thigh linear joint module 2-2 is placed inside the thigh structural frame 2-1, and its two ends are respectively connected to the root of the thigh structural frame 2-1 and the pivot of the root of the knee joint link 2-4 by hinges.
[0109] The output shaft of the thigh linear joint module 2-2 is connected to the thigh link 2-3, the thigh link 2-3 is hinged to the knee link 2-4, and the knee link 2-4 is hinged to the lower leg structural frame 2-5.
[0110] The thigh linear joint module 2-2 is installed inside the thigh structural frame and drives the thigh link 2-3 to achieve flexion and extension movements of the thigh relative to the knee joint. The linear drive mechanism facilitates larger output force and higher structural rigidity. The knee joint link 2-4 connects the thigh link 2-3 to the lower leg structural frame 2-5, forming the main kinematic pair of the knee joint. Through the combination of the linear joint module and the linkage mechanism, large-angle flexion and extension movements of the knee joint are achieved, while reducing the structural stress caused by direct load-bearing on the rotary joint.
[0111] The lower leg-ankle joint system, such as Figure 4 In Figure 4 (a) Figure 4 As shown in (b), it includes the medial lower leg linear joint module 3-1, the lateral lower leg linear joint module 3-2, the ankle joint base 3-3, the ankle joint universal joint 3-4, and the foot 3-5.
[0112] The inner lower leg linear joint module 3-1 and the outer lower leg linear joint module 3-2 are respectively arranged on both sides of the lower leg structural frame 2-5; the dual linear drive structure can realize coordinated control of ankle joint posture, while improving system redundancy and load-bearing capacity. The two ends of the inner lower leg linear joint module 3-1 and the outer lower leg linear joint module 3-2 are respectively connected to the axis of the lower leg structural frame 2-5 and the ankle joint base 3-3 by hinges.
[0113] The ankle joint base 3-3 receives signals from the linear drive module and connects the lower leg structure frame 2-5 to one end of the ankle joint universal joint 3-4. The ankle joint universal joint 3-4 provides the ankle joint with multi-degree-of-freedom rotation capability, enabling the foot to move in the pitch and roll directions. The other end of the ankle joint universal joint 3-4 is connected to the foot 3-5. The foot 3-5, as the end effector of the lower limb system, is used to contact the ground and transmit support and reaction forces.
[0114] The inner rotating joint module 1-2, the middle rotating joint module 1-3, the outer rotating joint module 1-4, the thigh linear joint module 2-2, the inner lower leg linear joint module 3-1, and the outer lower leg linear joint module 3-2 integrate motors, reducers, drivers, encoders, and force sensors inside the joints; the hip joint motor base 1-1 is fixedly mounted with an inertial measurement unit (IMU).
[0115] On the other hand, the present invention also provides a control method for the lower limb of a humanoid hip joint parallel drive robot, such as... Figure 8 As shown, the lower limb of the robot, driven by a parallel humanoid hip joint, is implemented as described above, specifically including:
[0116] Construct a kinematic model;
[0117] The hip joint parallel system, based on the characteristics of the human leg hip joint, adopts a coaxial drive layout, thus solving the problem of limited pitch angle. All rotational joint axes intersect at a single point, defined as the spherical rotation center, i.e., the origin of the mechanism. The included angle between the inner connecting rods 1-8 and the middle connecting rods 1-9 is... One end is connected to three sets of rotary joint modules respectively; the vector of the connected joint axis is defined as , This indicates three branches, with the other end hinged to the middle connecting rod 1-9, and the bending radius is... , The angle is represented by the semi-cone angle of the hip joint output flanges 1-11, and the connected joint axis vector is defined as... Together, they form a branch; the branches are spaced 120° counterclockwise, and the branches are connected to the hip joint output flange 1-11 via a revolute joint; the axis vector of the hip joint output flange 1-11 is defined as... The included angle between different rotational joint axes is ( Define the origin of the world coordinate system. Construct a three-dimensional orthogonal right-handed coordinate system at the center point of the hip joint output flange 1-11 plane. The shaft is perpendicular to the hip joint output flange 1-11. The axis is the negative direction of the joint axis vector. The axis is determined by the right-hand rule; for branches ( The corresponding mechanism parameters include: the structural angle of the drive linkage. , representing a unit vector and The included angle; the structural angle of the follower link. , representing a unit vector and The included angle, such as Figure 7 As shown; the semi-cone angle of the static platform , representing a unit vector and The included angle; projecting the static platform onto On the plane formed by the axes, the distribution angles of each branch after projection are represented by parameters. ( ) indicates that its meaning is Axis to Unit Vector Projection lines in the world coordinate system The angle between the two sides is defined as the positive direction of rotation, with the counterclockwise direction being the positive direction.
[0118] Using the aforementioned kinematic model, kinematic solutions are calculated, specifically including inverse kinematics and forward kinematics:
[0119] The inverse kinematics solution specifically involves: based on the given hip joint output flange 1-11 posture angles... , , Solve for the output angles of the three sets of rotary joint modules. The inverse kinematics can be obtained from the closed-loop condition of the mechanism's branches;
[0120] The inverse kinematics of the parallel mechanism and the unit vector containing the rotation axis of each branch. , , Related. Based on the coordinate relationship, the unit vector of each branch. ( ) is represented as:
[0121] ;
[0122] Unit vector ( The ) indicates the axis where the intermediate rotary joint is located, and the angle is output by the rotary joint module. have to:
[0123] ;
[0124] In the formula, Indicates about the axis Rotation angle The rotation transformation matrix, unit vector ( The symbol () indicates the axis containing the rotary joint connected to the hip joint output flange 1-11, which is related to the orientation of the hip joint output flange 1-11. Euler angles are used. Rotation transformation matrix of the sequence To describe the posture of hip joint output flanges 1-11:
[0125] ;
[0126] ;
[0127] ( This represents the unit vector of the axis of the rotational joint connected to the hip joint output flange 1-11 when the hip joint output flange 1-11 is in its initial pose. ;
[0128] ( The projection of the hip joint output flange 1-11 is denoted as ,in for:
[0129] ;
[0130] Rotation axis unit vector , The constraint of a closed-loop mechanism must be satisfied:
[0131] , ;
[0132] in According to the Law of Cosines for the sphere: ;
[0133] Taking the first link as an example, Substituting the constraints, we get: ;
[0134] ;
[0135] To obtain the output angle of the first branch rotary joint module ,make:
[0136] ;
[0137] We obtain the information used to describe the inverse motion of the first branch. The quadratic equation of :
[0138] ;
[0139] in:
[0140] ;
[0141] Solving the quadratic equation in one variable yields the input angle: ;
[0142] Similarly, for other branches, we obtain their quadratic equations and solve for the corresponding solutions. The output angle solutions for the three sets of revolute joint modules are as follows:
[0143] , ;
[0144] It can be seen that each branch may have at most two solutions, corresponding to two different configurations. Therefore, the inverse kinematics solution has at most eight solutions depending on the bias of the follower link. However, different solutions correspond to different branch configurations, so once the mechanism configuration is uniquely determined, the result of the inverse kinematics solution is also uniquely determined.
[0145] The kinematic solution specifically refers to: based on the angles output by the three sets of rotational joint modules. Solve for the attitude angle of hip joint output flange 1-11. The maximum number of forward kinematic solutions for a three-degree-of-freedom spherical parallel mechanism is eight, thus allowing for the derivation of the forward kinematic solutions.
[0146] First, establish a unit vector representing the rotation axis of each branch. in, The expression form corresponds to the inverse kinematics solution, while the expression for the unit vector is the same. When considering the overall integrity of the hip joint output flange 1-11, the coordinate transformation relationship should be adjusted accordingly. Represented as:
[0147] ;
[0148] Introducing unknowns , defined as the rotation angle of the middle rotation axis of the first branch, simplifies by substituting into the rotation matrix:
[0149] ;
[0150] Where vectors are used This represents a known parameter, specifically as follows:
[0151] , , ;
[0152] Therefore, the unit vector of the output axis of the second branch Represented as: ;
[0153] Substituting the rotation matrix and simplifying, we get: ;
[0154] Where vectors are used This represents a known parameter, specifically as follows:
[0155] , , ;
[0156] because They are all fixed to the hip joint output flange 1-11, so Decomposed into: ;
[0157] in For constant scalar, according to Calculation, that is: , , ;
[0158] Based on the constraints of institutional closed chains, establish and The two constraint equations in:
[0159] , ;
[0160] This leads to the information about the unknown quantity. The system of equations:
[0161] ;
[0162] in , , , , , ;
[0163] Next, we will discuss the unknown quantities. The system of equations can be transformed into the following form:
[0164] ;
[0165] according to , and then , ,in After simplification, we get The octet equation in one variable is: ;
[0166] Solve the octet equation in one variable to obtain the analytical expression for y, then: ;
[0167] Therefore, obtain ;
[0168] according to Inverse Euler angle solution, if: ;
[0169] but This is the correct kinematic solution.
[0170] The control method proposed in this invention employs a hybrid intelligent control architecture combining model control, reinforcement learning compensation, and state feedback closed-loop control for controlling the lower limb system of a humanoid robot. Based on a kinematic model and using reinforcement learning for compensation, this lower limb motion control method enables the robot to achieve stable walking, balance control, and dynamic posture adjustment under complex terrain conditions. The control system specifically includes:
[0171] (1) Gait and task generation module: Based on the robot's walking task (straight, turning, standing, going up and down slopes, etc.), it generates the expected trajectory of the foot in space and the reference motion of the robot's center of mass (CoM). This module is independent of the specific lower limb configuration and is the task space planning layer.
[0172] (2) Hybrid lower limb kinematics calculation module: For the hybrid configuration of the hip joint as a three-degree-of-freedom parallel mechanism, the ankle joint as a two-degree-of-freedom parallel mechanism, and the knee joint as a single-degree-of-freedom serial joint, this invention establishes a hierarchical kinematics calculation method: First, based on the desired pose of the foot and the position of the pelvis, the geometric configuration of the entire lower limb in space is calculated. Then, this geometric configuration is decomposed into three parts: hip posture, knee flexion, and ankle posture. For the hip and ankle joints, the inverse kinematics model is used to map the required spatial posture into the displacement or angle of each driving branch. For the knee joint, the corresponding joint angle is directly obtained through the geometric relationship between the knee and ankle joints. Through this decomposition and mapping method, a unified conversion from "foot and body target" to "each driving target" is achieved.
[0173] (3) Cooperative Control Module: After completing the kinematic calculation, the system obtains the target position, velocity, or torque of each drive actuator. These targets constitute the model-based primary control commands. This control layer mainly generates executable motion commands based on the geometric model and structural constraints of the lower limbs to ensure that the geometric closed-loop constraints of the parallel mechanism are not violated and that the overall posture of the lower limbs meets the planned gait and foot trajectory requirements. This layer of control can ensure that the robot can complete basic walking and supporting actions under ideal model conditions, but it is still difficult to cope with uneven ground, impact disturbances, and model errors in the real environment.
[0174] (4) To improve the robustness and adaptability of lower limb movement, this invention introduces a reinforcement learning control module on the basis of model-based control. The module takes the real-time state of the robot's lower limbs as input, including joint angles, joint velocities, body posture, foot contact state, and ground forces. Through reinforcement learning, it outputs compensation quantities for the control of each joint, correcting errors caused by the kinematic and dynamic models, enabling the robot to maintain stable walking and posture balance even in uncertain terrain or under external disturbances. The reinforcement learning strategy can be trained offline in a simulation environment and run online in an actual robot.
[0175] The reinforcement learning formula is as follows: ;
[0176] Where u represents the system control input, including the control inputs for the three rotary motors of the hip joint and the three linear motors of the knee and ankle joints. For stabilizing control items, To enhance learning output, Adaptive weights related to structural risk;
[0177] In this invention, kinematics calculation and reinforcement learning are not mutually exclusive, but rather work together. The kinematic model is responsible for providing basic motion commands that conform to the mechanism constraints, ensuring that the parallel mechanism does not experience structural conflicts or infeasible movements; reinforcement learning control, on this basis, compensates for execution errors, ground changes, and dynamic uncertainties, and superimposes them to form the final control commands, enabling the robot to maintain structural safety while possessing good environmental adaptability.
[0178] (5) Low-level execution and closed-loop feedback module: The final control command is sent to the inner rotary joint module 1-2, the middle rotary joint module 1-3 and the outer rotary joint module 1-4, the thigh linear joint module 2-2, the inner lower leg linear joint module 3-1 and the outer lower leg linear joint module 3-2. The module outputs torque or displacement according to the control command, so that the robot's lower limbs produce corresponding movements. At the same time, sensor information is continuously collected and the feedback information is sent to the reinforcement learning control module and the serial-parallel hybrid lower limb kinematics calculation module, thereby forming a closed-loop control system, so that the lower limb movement can be corrected in real time and run stably.
[0179] Without departing from the concept of this invention:
[0180] The drive unit can be a motor, a hydraulic actuator, a pneumatic actuator, or a combination thereof.
[0181] The parallel transmission chain can adopt various equivalent spherical parallel mechanism forms;
[0182] The triaxial coaxial arrangement can be completely collinear or approximately collinear within the manufacturing allowable range;
[0183] The hip joint mechanism can integrate a torque sensor, a position sensor, or an encoder to achieve closed-loop control;
[0184] This invention is also applicable to shoulder joints, spinal joints, or other multi-degree-of-freedom load-bearing joints.
[0185] In this embodiment, a humanoid hip joint parallel drive robot performs flexion and extension movements on its lower limbs, such as... Figure 5 As shown, where Figure 5 (a) represents the highest point of the flexion-extension movement. Figure 5(b) is the position after flexion and extension; rotation and abduction movements are as follows: Figure 6 As shown, where Figure 6 (a) represents the highest point of the flexion-extension movement. Figure 6 (b) is the position after the flexion and extension movement.
[0186] Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or part of the technical solution, can be embodied in the form of a computer program product.
[0187] The various embodiments in this application are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.
[0188] The scope of protection of this application is not limited to the embodiments described above. Obviously, those skilled in the art can make various modifications and variations to this disclosure without departing from the scope and spirit of this disclosure. If such modifications and variations fall within the scope of the methods disclosed herein and their equivalents, then the intent of this disclosure also includes such modifications and variations.
Claims
1. A humanoid hip joint parallel-driven robot lower limb, characterized in that, This includes the parallel hip joint system, the thigh-knee joint system, and the lower leg-ankle joint system; The parallel hip joint system includes a hip joint motor base, a rotary joint module, an inner output link, a middle output link, an outer output link, an inner connecting link, a middle connecting link, an outer connecting link, a hip joint output flange, a bearing, a bearing end cover, and a bearing central shaft, forming a three-chain parallel mechanism. The thigh-knee joint system includes a thigh structural frame, a thigh linear joint module, a thigh link, a knee link, and a lower leg structural frame. The thigh structural frame is a hollow structure, connected to the hip joint output flange. A linear drive module is installed inside the thigh structural frame and transmits power to the lower leg structural frame. The thigh linear joint module is placed inside the thigh structural frame, and its two ends are respectively connected to the root of the thigh structural frame and the root of the knee joint connecting rod at the pivot point. The output shaft of the thigh linear joint module is connected to the thigh connecting rod, the thigh connecting rod is hinged to the knee joint connecting rod, and the knee joint connecting rod is hinged to the lower leg structural frame. The lower leg-ankle joint system includes a medial lower leg linear joint module, a lateral lower leg linear joint module, an ankle joint base, an ankle joint universal joint, and a foot. The inner lower leg linear joint module and the outer lower leg linear joint module are respectively arranged on both sides of the lower leg structural frame; the two ends of the inner lower leg linear joint module and the outer lower leg linear joint module are respectively connected to the axis of the lower leg structural frame and the ankle joint base by hinges, and the other end of the ankle joint universal joint is connected to the foot. The rotary joint module has three sets, which are fixed to the hip joint motor base. The output shafts of the three sets of rotary joint modules are respectively connected to one end of the inner output link, the middle output link and the outer output link. The other ends of the inner output link, the middle output link and the outer output link are respectively connected to one end of the inner connecting link, the middle connecting link and the outer connecting link through bearings. The other ends of the inner connecting link, the middle connecting link and the outer connecting link are connected to the hip joint output flange through bearings. The rotary joint module includes an inner rotary joint module, a middle rotary joint module, and an outer rotary joint module, each containing an independent drive motor and a reduction mechanism; the inner output link, the middle output link, and the outer output link are respectively connected to the inner rotary joint module, the middle rotary joint module, and the outer rotary joint module. The bearing, bearing end cap, and bearing central shaft together form a rotational hinge for the drive linkage, which is connected between the outer connecting linkage and the hip joint output flange to achieve hinge connection.
2. The lower limb of a humanoid hip joint parallel-driven robot according to claim 1, characterized in that, The ankle joint base receives signals from the linear drive module and connects the lower leg structure frame to one end of the ankle joint universal joint. The linear drive module achieves precise linear motion servoing through motor movement and feedback control.
3. The lower limb of a humanoid hip joint parallel-driven robot according to claim 2, characterized in that, The inner rotary joint module, middle rotary joint module, outer rotary joint module, thigh linear joint module, inner calf linear joint module, and outer calf linear joint module integrate motors, reducers, drivers, encoders, and force sensors inside the joints; the hip joint motor base is fixedly mounted with an inertial measurement unit (IMU).
4. The lower limb of a humanoid hip joint parallel-driven robot according to claim 3, characterized in that, The humanoid hip joint parallel drive robot lower limb also includes a control system for controlling the humanoid robot lower limb system; The control system specifically includes: (1) Gait and task generation module: Based on the robot's walking task, generate the expected trajectory of the foot in space and the reference motion of the robot's center of mass; (2) Series-parallel hybrid lower limb kinematics solution module: A hierarchical kinematics solution method was established: First, based on the desired pose of the foot and the position of the pelvis, the geometric configuration of the entire lower limb in space was calculated. Then, the geometric configuration was decomposed into three parts: hip posture, knee flexion, and ankle posture. For the hip and ankle joints, the inverse kinematics model was used to map the required spatial posture into the displacement or angle of each driving branch. For the knee joint, the corresponding joint angle was obtained directly through the geometric relationship between the knee and ankle joints. (3) Cooperative control module: obtains the target position, velocity or torque of each drive actuator, generates executable motion commands based on the geometric model and structural constraints of the lower limb, and ensures that the overall posture of the lower limb meets the planned gait and foot trajectory requirements; (4) Reinforcement learning control module: Taking the real-time state of the robot's lower limbs as input, it outputs compensation quantities for the control quantities of each joint through reinforcement learning, correcting the errors caused by the kinematic model and dynamic model; The reinforcement learning formula is as follows: ; Where u represents the system control input, including the control inputs for the three rotary motors of the hip joint and the three linear motors of the knee and ankle joints. For stabilizing control items, To enhance learning output, Adaptive weights related to structural risk; The kinematic model is responsible for providing basic motion commands that conform to the mechanism constraints. Based on this, compensation is made for execution errors, ground changes and dynamic uncertainties, and the final control commands are formed by superimposing them. (5) Low-level execution and closed-loop feedback module: The final control command is sent to the inner rotary joint module, the middle rotary joint module, the outer rotary joint module, the thigh linear joint module, the inner lower leg linear joint module, and the outer lower leg linear joint module. The module outputs torque or displacement according to the control command, so that the robot's lower limbs produce corresponding movements. At the same time, sensor information is continuously collected and the feedback information is sent to the reinforcement learning control module and the serial-parallel hybrid lower limb kinematics calculation module, thereby forming a closed-loop control system, so that the lower limb movement can be corrected in real time and run stably.
5. A control method for the lower limb of a humanoid hip joint parallel-driven robot, implemented using the humanoid hip joint parallel-driven robot lower limb described in claim 4, characterized in that... Specifically, it includes: Construct a kinematic model; Kinematic solutions are performed using the aforementioned kinematic model.
6. The control method for the lower limb of a humanoid hip joint parallel drive robot according to claim 5, characterized in that, The kinematic model is as follows: The hip joint parallel system, based on the characteristics of the human leg hip joint, adopts a coaxial drive layout to address the limitation of pitch angle. All rotational joint axes intersect at a single point, defined as the spherical rotation center, i.e., the origin of the mechanism. The angle between the inner connecting rod and the middle connecting rod is... One end is connected to three sets of rotary joint modules; the vector of the connected joint axis is defined as follows: , This indicates three branches, with the other end hinged to the middle connecting rod, and the curvature is... , The angle is represented by the half-cone angle of the hip joint output flange, and the connected joint axis vector is defined as... Together, they form a branch; the branches are spaced 120° counterclockwise, and the branches are connected to the hip joint output flange via a revolute joint; the axis vector of the hip joint output flange shaft is defined as... The included angle between different rotational joint axes is , Define the origin of the world coordinate system. Construct a three-dimensional orthogonal right-handed coordinate system at the center point of the hip joint output flange plane. The shaft is perpendicular to the hip joint output flange. The axis is the negative direction of the joint axis vector. The axis is determined by the right-hand rule; for branches The corresponding mechanism parameters include: the structural angle of the drive linkage. , representing a unit vector and The included angle; the structural angle of the follower link. , representing a unit vector and The included angle; the semi-cone angle of the static platform , representing a unit vector and The included angle; projecting the static platform onto On the plane formed by the axes, the distribution angles of each branch after projection are represented by parameters. It means that Axis to Unit Vector Projection lines in the world coordinate system The angle between the two sides is defined as the positive direction of rotation, with the counterclockwise direction being the positive direction.
7. The control method for the lower limb of a humanoid hip joint parallel drive robot according to claim 6, characterized in that, The kinematic solution specifically includes inverse kinematics and forward kinematics: The inverse kinematics solution specifically involves: based on the given hip joint output flange posture angle... , , Solve the output angle of the rotary joint module The inverse kinematics can be obtained from the closed-loop condition of the mechanism's branches; Unit vector of each branch Represented as: ; Unit vector , indicating the axis where the intermediate rotational joint is located. The angle is output by the rotary joint module. have to: ; In the formula, Indicates about the axis Rotation angle The rotation transformation matrix, unit vector The axis containing the rotary joint connected to the hip joint output flange is indicated using Euler angles. Rotation transformation matrix of the sequence To describe the hip joint output flange orientation: ; ; This represents the unit vector of the axis of the rotational joint connected to the hip joint output flange when the hip joint output flange is in its initial pose. ; The projection of the hip joint output flange is denoted as ,in for: ; Rotation axis unit vector , The constraint of a closed-loop mechanism must be satisfied: , ; in According to the Law of Cosines for the sphere: ; The output angle solution of the three sets of rotary joint modules is expressed as follows: , ; The kinematic solution specifically refers to: based on the angles output by the three sets of rotational joint modules. Solve for the three-axis attitude angles of the hip joint output flange. The maximum number of forward kinematic solutions for a three-degree-of-freedom spherical parallel mechanism is eight, thus allowing us to derive the forward kinematic solutions. First, establish a unit vector representing the rotation axis of each branch. in, The expression form corresponds to the inverse kinematics solution, while the expression for the unit vector is the same. When considering the overall integrity of the hip joint output flange, the coordinate transformation relationship should be adjusted accordingly. Represented as: ; Introducing unknowns , defined as the rotation angle of the middle rotation axis of the first branch, simplifies by substituting into the rotation matrix: ; Where vectors are used This represents a known parameter, specifically as follows: , , ; Therefore, the unit vector of the output axis of the second branch Represented as: ; Substituting the rotation matrix and simplifying, we get: ; Where vectors are used This represents a known parameter, specifically as follows: , , ; because They are all fixed to the hip joint output flange, so Decomposed into: ; in For constant scalar, according to Calculation, that is: , , ; Based on the constraints of institutional closed chains, establish and The two constraint equations in: , ; This leads to the information about the unknown quantity. The system of equations: ; in , , , , , ; Next, we will discuss the unknown quantities. The system of equations can be transformed into the following form: ; according to , and then , ,in After simplification, we get The octet equation in one variable is: ; Solve the octet equation in one variable to obtain the analytical expression for y, then: ; Therefore, obtain ; according to Inverse Euler angle solution, if: ; but This is the correct kinematic solution.
Citation Information
Patent Citations
Wearable and size-adjustable steel wire rope driven lower limb exoskeleton robot
CN118340648A
Modularized flexible lower limb exoskeleton robot
CN222942618U