Robotic leg structure and method of controlling the same

Abstract

The application discloses a robot leg structure and a control method thereof. The structure comprises a hip joint module, a thigh module, a knee joint module, a shank module and a foot end module. A flat plate stopper is arranged on a structure connecting surface of each of the hip joint module, the thigh module, the knee joint module, the shank module and the foot end module, and the modules are connected based on the flat plate stopper. The hip joint module comprises a first driving unit, a second driving unit and a third driving unit. The first driving unit controls a roll freedom degree, the second driving unit controls a pitch freedom degree, and the third driving unit controls a yaw freedom degree. Driving axes of the first driving unit, the second driving unit and the third driving unit are perpendicular to each other and intersect at the same point. The knee joint module is driven based on a torque motor. A rotor of the torque motor is connected with the thigh module, and a stator of the torque motor is connected with the shank module. The robot leg structure has the advantages of improving motion stability, enhancing joint control precision and optimizing dynamic response capability.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of robots, in particular to a robot leg structure and a control method thereof. BACKGROUND

[0002] As an industry chain and a high-end technology chain in the field of robot technology, the lower limb structure design of a humanoid robot directly determines the motion stability, environmental adaptability and dynamic response capability of the robot. As a core component of the lower limb, the design of the hip joint, the knee joint and the foot end directly affects the overall performance of the robot.

[0003] At present, the traditional humanoid robot has many problems in design. In terms of structural design, for example, the traditional humanoid robot hip joint usually adopts a series configuration or a parallel configuration to realize the movement of three degrees of freedom. However, these configuration methods have the technical defects of serious motion interference and difficulty in accurately converging the axes of the joints at a point, which leads to significant kinematic coupling and increases the complexity of dynamic calculation. For another example, the traditional humanoid knee joint adopts a four-bar mechanism or a similar principle for motion transmission, which limits the movement angle of the thigh and makes it difficult to complete the action of large-scale flexion. For another example, the traditional foot is assembled by multiple parts, and the cumulative tolerance of the parts makes it difficult to ensure the assembly precision. Under the action of long-term dynamic load, the connecting parts are prone to micron-level wear, gradually expanding the fitting gap, and eventually leading to the shaking phenomenon during movement. In terms of structural connection, the leg structure of the traditional humanoid robot usually adopts pin positioning and screw fastening, which has the problems of difficult installation precision, easy wear leading to increased fitting gap, etc. At the same time, the screw bears shear force and bending moment, and is prone to fatigue fracture or loosening failure under the action of long-term alternating load.

[0004] Based on the above problems, the humanoid robot not only has low overall performance, but also has the problems of poor maintenance convenience and time-consuming and labor-consuming, which further affects the use efficiency. SUMMARY

[0005] The purpose of the embodiments of the present application is to provide a robot leg structure and a control method thereof, which have the advantages of improving motion stability, enhancing joint control precision and optimizing dynamic response capability.

[0006] In a first aspect, the embodiments of the present application provide a robot leg structure, comprising: a hip joint module, a thigh module, a knee joint module, a calf module and a foot end module, wherein the hip joint module, the thigh module, the knee joint module, the calf module and the foot end module are all provided with a flat stop on a structural connection surface and are connected based on the flat stop.

[0007] The hip joint module comprises a first driving unit, a second driving unit and a third driving unit, the first driving unit controls the roll degree of freedom, the second driving unit controls the pitch degree of freedom, and the third driving unit controls the yaw degree of freedom, and driving axes of the first driving unit, the second driving unit and the third driving unit are perpendicular to each other and intersect at the same point.

[0008] The knee joint module is driven based on a torque motor, and a rotor of the torque motor is connected with the thigh module, and a stator of the torque motor is connected with the shank module.

[0009] In a second aspect, the embodiments of the present application provide a control method of a robot leg structure, comprising:

[0010] The captured motion target is decomposed into an expression layer target and a motion layer target, the expression layer target is used to constrain leg joint positions and key point positions, and the motion layer target is used to control linear velocities, attitude angles and heights of a whole body root node of the robot;

[0011] The expression layer target and the motion layer target are analyzed by using the trained strategy model, and control instructions of a target robot are output.

[0012] The embodiments of the present application provide a robot leg structure and a control method thereof, the structure comprising a hip joint module, a thigh module, a knee joint module, a shank module and a foot end module, and flat plate stop mouths are arranged on structure connection surfaces of the hip joint module, the thigh module, the knee joint module, the shank module and the foot end module, and the flat plate stop mouths are connected based on the flat plate stop mouths, through three-degree-of-freedom orthogonal intersection hip joint driving design, torque motor direct drive knee joint and modular flat plate stop mouth connection structure, the problems of motion coupling, transmission precision loss and insufficient assembly precision of a traditional robot lower limb are effectively solved, and the robot leg structure has the advantages of improving motion stability, enhancing joint control precision and optimizing dynamic response capability. BRIEF DESCRIPTION OF DRAWINGS

[0013] Figure 1 is a structure schematic diagram of the robot leg structure provided by the embodiments of the present application;

[0014] Figure 2 is a cross-sectional schematic diagram of the hip joint module provided by the embodiments of the present application;

[0015] Figure 3 is a side structure schematic diagram of the robot leg structure provided by the embodiments of the present application;

[0016] Figure 4 is a structure schematic diagram of the spline structure provided by the embodiments of the present application;

[0017] Figure 5is a structural schematic diagram of a flat end collar and spline structure provided by an embodiment of the present application.

[0018] Figure 6 is a structural schematic diagram of a foot end module provided by an embodiment of the present application.

[0019] Figure 7 is a structural schematic diagram of an elastic foot bottom provided by an embodiment of the present application.

[0020] In the figure, 1 is a hip joint module, 2 is a thigh module, 3 is a knee joint module, 4 is a lower leg module, and 5 is a foot end module. DETAILED DESCRIPTION

[0021] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative labor fall within the scope of protection of the present application.

[0022] It should be understood that each step described in the method embodiments disclosed in the present application can be executed in different orders and / or in parallel. In addition, the method embodiments can include additional steps and / or omit the execution of the steps shown. The scope of the present application is not limited in this respect.

[0023] The term "comprising" and variations thereof as used in the present application are open-ended, that is, "including but not limited to". The term "based on" is "at least partially based on". The term "one embodiment" means "at least one embodiment"; the term "another embodiment" means "at least one additional embodiment"; the term "some embodiments" means "at least some embodiments". Related definitions of other terms will be given in the following description.

[0024] In related technologies, the design of the lower limb structure of a humanoid robot faces technical bottlenecks such as serious coupling of hip joint kinematics, loss of precision caused by a long transmission chain of the knee joint, and difficulty in balancing the weight and stiffness of the foot end. The traditional hip joint configuration has an axis intersection deviation that causes an inertia accumulation problem, the knee joint drive system has energy loss due to too many transmission links, and the assembly precision of the foot is affected by the cumulative effect of part tolerances, resulting in an increase in the gap. In a certain warehouse logistics scenario, the robot needs to perform high-frequency steering and precise parking actions among complex shelves, but the existing structure cannot meet the millimeter-level positioning requirements due to joint coupling and foot end gap, resulting in a deviation of the motion trajectory.

[0025] To solve the technical problems in the related art, please refer to Figures 1 to 7 is a structural schematic diagram of the overall structure of a robot leg and related components provided by an embodiment of the present application.

[0026] In an embodiment, the robot leg structure comprises a hip joint module 1, a thigh module 2, a knee joint module 3, a lower leg module 4 and a foot end module 5, and a flat plate stop is arranged on the structure connecting surface of the hip joint module 1, the thigh module 2, the knee joint module 3, the lower leg module 4 and the foot end module 5, and the modules are connected based on the flat plate stop; the hip joint module 1 comprises a first driving unit, a second driving unit and a third driving unit, the first driving unit controls the roll degree of freedom, the second driving unit controls the pitch degree of freedom, and the third driving unit controls the yaw degree of freedom, the driving axes of the first driving unit, the second driving unit and the third driving unit are perpendicular to each other and intersect at the same point; the knee joint module 3 is driven based on a torque motor, and the rotor of the torque motor is connected with the thigh module, and the stator of the torque motor is connected with the lower leg module.

[0027] The flat plate stop refers to a planar flange arranged on the structure connecting surface, which can form a continuous annular plane by machining, and the modules are positioned by surface contact. The intersection of the axes of the three driving units refers to the spatial geometric relationship of the driving shafts, which can realize orthogonal arrangement of the axes by a frame structure, and the intersection point can be determined by three-dimensional coordinate calibration, specifically the hip joint center of motion. The torque motor direct drive refers to that the electromagnetic assembly directly transmits torque, which can use a frameless torque motor, the rotor is connected with the thigh by a shrink fit process, and the stator is connected with the lower leg by an interference fit.

[0028] In an embodiment, referring to Figure 1 The robot leg structure comprises five modules, i.e. a hip joint module 1, a thigh module 2, a knee joint module 3, a lower leg module 4 and a foot end module 5, and the combination of the flat plate stop machining process and the spline connection mode improves the motion performance, structural reliability and maintainability of the robot. The modules are assembled and disassembled by the standardized flat plate stop, and the motor is installed and fixed by the spline connection mode when the motor is installed.

[0029] Specifically, the hip joint module 1 forms a spatial spherical pair by orthogonal arrangement of three driving units, the intersection point of the three driving shafts constitutes the kinematic origin, and the inertia coupling of the multi-degree-of-freedom motion is eliminated. The torque motor of the knee joint module 3 directly connects the thigh and the lower leg, the electromagnetic torque is directly transmitted through the air gap, and the elastic deformation of the mechanical transmission chain is avoided. The modules are axially positioned by the precisely machined flat plate stop, and the assembly cumulative error is eliminated by surface contact. When a turning action is performed, the three-axis coordinated motion of the hip joint generates a pure rotation torque, the torque motor used in the knee joint module 3 adjusts the flexion angle in real time according to the ground reaction force, and the flat plate stop structure between the modules effectively suppresses vibration transmission.

[0030] Here, the pure rotation movement is achieved by orthogonal intersection driving to complete the movement of the hip joint 1, and the movement adjustment is achieved by using the direct driving mode in the knee joint module 3. The control deviation caused by the coupling of the hip joint movement is effectively solved, and the precision loss caused by the transmission chain of the knee joint is eliminated, and the connection mode of the flat plate stop port also effectively suppresses the gap accumulation between the connection surfaces of each module.

[0031] Further, the output shaft of the first driving unit is fixedly connected with the pitch frame, and the output shaft of the second driving unit is fixedly connected with the yaw frame.

[0032] The pitch frame refers to a rigid frame structure supporting the pitch degree of freedom movement, which can be realized by aluminum alloy casting or carbon fiber composite material molding, and the inner wall of the frame is provided with a mounting interface matched with the output end of the driving unit. The frame serves as a load-bearing main body of the pitch movement, and by directly connecting the output end of the driving unit, the intermediate link in the traditional transmission chain is eliminated. The yaw frame refers to a ring-shaped support structure supporting the yaw degree of freedom movement. The frame forms a motion transmission path with the output end of the driving unit through rigid connection, ensuring the precise intersection of the yaw axis and other degree of freedom axes.

[0033] Specifically, the output end of each driving unit can be physically connected to the corresponding frame through a spline or flange interface. The rotation center axis of the pitch frame coincides with the output axis of the first driving unit. When the first driving unit works, it directly drives the pitch frame to rotate around its axis to realize the pitch movement. The mounting plane of the yaw frame is perpendicular to the pitch frame axis, and the output torque of the second driving unit directly acts on the yaw frame through rigid connection, making it rotate around the preset axis to complete the yaw action. The intersection point of the axes of the two frames and the axis of the third driving unit form a spatial orthogonal relationship, realizing the movement decoupling of the three degrees of freedom.

[0034] In the above embodiment, by the frame direct connection mode, the driving axes are geometrically intersected in the physical space, eliminating the movement interference caused by axis deviation, making the rotation centers of the three degrees of freedom accurately intersect at a single point, and reducing the movement control complexity and control accuracy.

[0035] Further, the first driving unit and the second driving unit adopt a planetary reducer quasi-direct-drive motor, and the third driving unit adopts a cycloidal reducer quasi-direct-drive motor, and the reduction ratio of the cycloidal reducer is 1:20.

[0036] Specifically, in the hip joint module 1, the adjustment and motion control of the robot posture are realized based on the cooperative work of the first driving unit, the second driving unit and the third driving unit. Among them, the first driving unit and the second driving unit adopt a planetary reducer to drive the roll and pitch freedom degrees, eliminating the reverse clearance and transmission error brought by the traditional reducer, and the third driving unit adopts a cycloid reducer to drive the motor, which converts the motor speed into low-speed large-torque output by using its high reduction ratio, thereby meeting the load demand of the yaw freedom degree. The combination of the three driving units balances the motion precision and load capacity while keeping the structure compact.

[0037] Further, the knee joint module 3 adopts a cross-roller bearing as a main support structure.

[0038] Among them, the cross-roller bearing refers to a rolling bearing in which two groups of rollers are arranged in an orthogonal manner. It can be realized by adopting a structure with V-shaped grooves on the inner and outer rings, and the rollers and grooves are in line contact. This structure can simultaneously withstand radial load, axial load and overturning moment, and effectively disperses the complex load generated by joint movement through the cross-arrangement of the rollers to form bidirectional support.

[0039] Specifically, during the assembly of the knee joint module 3, the inner ring of the cross-roller bearing is rigidly connected to the end flange of the thigh module 2, and the outer ring is fixed to the proximal flange of the lower leg module 4. When the torque motor drives the knee joint module 3 to perform flexion and extension movement, the cross-roller bearing transmits the load through the rolling contact between the rollers and the grooves, and the orthogonally arranged roller groups bear the radial force in the rotation plane and the axial force perpendicular to the rotation axis, respectively. The internal pre-tightening force of the bearing is set by adjusting the tightening torque of the end cap bolt to ensure that there is no axial movement during movement. The geometric characteristics of the roller and groove contact line make the contact stress distribution uniform when the bearing bears the torque load, avoiding plastic deformation caused by local overload.

[0040] Further, the knee joint module 3 is integrated with a non-backlash brake.

[0041] Among them, the non-backlash brake refers to a brake device that eliminates the clearance between the motion pairs. It can be realized by using an electromagnetic brake or a mechanical brake, and the brake surface and the transmission component form a gapless contact. This device eliminates the reverse clearance in the transmission chain by applying a pre-tightening force, and immediately locks the output shaft position when the motor stops.

[0042] Specifically, the non-backlash brake is directly integrated on the torque motor shaft of the knee joint module 3. When the torque motor stops power supply, the friction plate of the brake is in contact with the brake disc under the action of spring pre-tightening force, forming a rigid connection. During movement, the brake remains in a pre-tightened state but does not interfere with normal rotation, and only when position locking is required is the brake disc fully pressed. The working state of the brake is switched through the feedback signal of the motor controller, ensuring that the braking action is synchronized with the motor operating state.

[0043] Further, the foot end module 5 is composed of a foot shell and a flexible foot bottom piece fixed connection, the foot shell is integrally formed by 3D printing of titanium alloy, and the inside of the foot shell is a dot matrix structure, and the flexible foot bottom piece is composed of a super-elastic polymer material.

[0044] Titanium alloy integrally formed by 3D printing refers to the use of additive manufacturing technology to melt and accumulate titanium alloy powder layer by layer to form a whole structure, which can be realized by laser selective melting process. This process can directly generate complex internal structures without subsequent mechanical processing. The dot matrix structure refers to a three-dimensional space grid composed of periodically arranged truss units, which can be realized by octahedral lattice units, which can achieve lightweight design while ensuring load-bearing capacity. Super-elastic polymer material refers to a polymer with high elasticity and energy absorption characteristics, which can be realized by polyurethane elastomer, and its elastic modulus can be designed according to the foot bottom contact pressure distribution. Specifically, the foot shell formed by 3D printing of titanium alloy eliminates the tolerance accumulation caused by traditional multi-part assembly through integrated manufacturing process, and the internal dot matrix structure effectively reduces the overall weight while maintaining the stiffness of the foot. The fixed connection of the flexible foot bottom piece and the shell forms a rigid-flexible composite structure, and the super-elastic polymer material absorbs impact energy by deforming when touching the ground, and quickly recovers its shape when leaving the ground. The anisotropic stiffness distribution of the dot matrix structure matches the direction of the foot contact force, maximizing the structural load efficiency.

[0045] Further, the foot end module 5 is composed of a foot shell and a flexible foot bottom piece fixed connection, the foot shell is integrally formed by 3D printing of titanium alloy, and the inside of the foot shell is a dot matrix structure, and the flexible foot bottom piece is composed of a super-elastic polymer material.

[0046] The multi-dimension force sensor refers to a sensing device capable of detecting forces in three orthogonal directions simultaneously, which can be implemented by using a piezoelectric sensor or a strain sensor, and is used for acquiring three-dimensional contact force data between the foot and the contact surface in real time. The pressure sensing matrix refers to a detection unit arranged in an array form by a plurality of pressure sensors, which can be implemented by using a flexible piezoresistive material or a capacitive sensor, and is used for acquiring the plantar pressure distribution state. The micro variable form accommodating cavity refers to a closed cavity structure with elastic deformation capability, which can be manufactured by using a silica gel material through a molding process, and the volume change is achieved by air pressure adjustment. The air pipe component refers to a passage assembly connecting the variable form accommodating cavity and an external air source, which can be implemented by using a polyurethane hose matched with a quick connector, and is used for transmitting compressed air or negative pressure air flow.

[0047] Specifically, the multi-dimension force sensor arranged inside the foot end module 5 is embedded between the foot shell and the elastic plantar component, and the force transmission path is isolated by a rigid mounting base. The pressure sensing matrix is integrated on the upper surface of the elastic plantar component, and the signal transmission is achieved by using a flexible circuit board. The micro variable form accommodating cavities are uniformly distributed inside the elastic plantar component, and each accommodating cavity is connected to an external air pump control system through an independent air pipe component. When the robot walks, the multi-dimension force sensor feeds back the ground reaction force vector in real time, and the pressure sensing matrix synchronously monitors the pressure gradient distribution of the plantar contact area. The external air pump dynamically adjusts the air pressure of each accommodating cavity according to the sensor data, so that the elastic plantar component generates local deformation, thereby changing the geometry of the plantar contact surface.

[0048] Further, the height of the flat shoulder is 1.5 mm, and the fit tolerance is H6 / g5.

[0049] The height of the flat shoulder refers to the axial dimension of the raised structure perpendicular to the connecting surface, which can be implemented by using precision milling processing. The fit tolerance H6 / g5 refers to the clearance fit level between the hole and the shaft, which can be implemented by using numerical control grinding process. H6 represents that the tolerance band of the hole is 6-level accuracy of the base hole, and g5 represents that the tolerance band of the shaft is 5-level accuracy of the base shaft, so as to form a precise clearance fit relationship.

[0050] Specifically, the flat shoulder with a height of 1.5 mm is arranged on the connecting surface of each module, and the axial positioning is achieved by the raised structure. When the modules are assembled, the side surface of the flat shoulder forms surface contact constraint with the corresponding groove, and the single-sided clearance is controlled in the micron level range by the fit H6 / g5 tolerance. Under the action of dynamic load, the height of the shoulder can effectively resist the deformation caused by the bending moment, and the tolerance fit relationship allows a small amount of thermal expansion without generating assembly stress.

[0051] Further, the first driving unit, the second driving unit, the third driving unit and the torque motor are fixed based on the spline structure, the modulus of the spline is 0.5, the pressure angle is 30°, and the matching gap is less than 0.01 mm.

[0052] The spline structure refers to a mechanical connection mode for realizing power transmission through the concave-convex tooth shape arranged on the inner and outer shaft surfaces, can be realized by using involute tooth shape or rectangular tooth shape, and realizes torque transmission and axis positioning through tooth surface contact. The modulus of 0.5 refers to a basic size parameter of the gear, can be realized by using a standard modulus series, and the modulus value determines the geometric proportional relationship of the tooth height and the pitch, directly affects the bearing capacity and structural compactness. The pressure angle of 30° refers to the included angle formed by the normal line of the tooth surface contact point and the tangent of the pitch circle, can be realized by adjusting the machining angle of the cutter, and the parameter affects the tooth surface contact stress distribution and bending strength. The matching gap is less than 0.01 mm, which refers to the radial gap control range of the spline pair after assembly, can be realized by using precision grinding and process selection, and ensures the transmission accuracy and motion stability.

[0053] Specifically, the output end of each driving unit and the connection part of the corresponding frame are machined with an internal spline tooth shape, and the outer surface of the motor shaft is machined with a matching external spline tooth shape. During assembly, the spline pair is engaged by axial pressing, and torque transmission and axis positioning are completed by tooth surface contact. The modulus of 0.5 can control the overall structure size under the premise of ensuring sufficient tooth height, and the pressure angle of 30° can balance the tooth surface contact stress and bending strength requirement. The matching gap is controlled within 0.01 mm through precision machining, avoiding transmission error or vibration caused by excessive gap.

[0054] It should be noted that, in addition to the first driving unit, the second driving unit, the third driving unit and the torque motor described, other motors or driving units on the robot leg structure that are not mentioned are connected and fixed by using the spline connection mode.

[0055] Further, a control method of a robot leg structure, comprising:

[0056] The captured moving target is decomposed into expression layer targets and motion layer targets, the expression layer targets are used to constrain the leg joint positions and key point positions, and the motion layer targets are used to control the linear velocity, attitude angle and height of the whole body root node of the robot;

[0057] The trained strategy model is used to analyze and process the expression layer targets and the motion layer targets, and control instructions of the target robot are output.

[0058] Wherein, the expression layer target refers to a motion constraint framework formed by defining joint angles and key point spatial coordinates, which can be implemented by using inverse kinematics algorithm or trajectory planner, for ensuring that the leg movement conforms to the biomechanical law, the motion layer target refers to the dynamic control requirement for the robot center of mass motion trajectory and attitude stability, which can be implemented by using model predictive control or optimization algorithm, for maintaining the overall motion balance of the robot, the strategy model refers to a multi-modal data processing module trained by reinforcement learning or deep learning, which can be implemented by using deep neural network or decision tree model, for fusing sensor data and generating real-time control instructions to control the motion of the robot.

[0059] Specifically, when the robot performs dynamic motion, the motion target is first split into joint-level position constraints and whole-body motion dynamics parameters. The expression layer target generates reference angles for each joint through inverse kinematics calculation, ensuring that the leg motion trajectory conforms to the preset gait characteristics; the motion layer target calculates the motion parameters of the root node through optimization algorithm, keeping the center of mass trajectory stable. The trained strategy model receives environmental perception data and target parameters, and generates motor torque instructions and joint angle adjustment signals through multi-level feature fusion, achieving precise motion control.

[0060] Exemplarily, when controlling the leg structure of the robot, it is in the case of controlling the target robot, by determining the current motion target, and then analyzing and processing the motion target, combining the pre-trained model to output the control result, to realize the control of the target robot.

[0061] When the motion target is decomposed, the expression layer target and the motion layer target are obtained, wherein the expression layer target : only constrains 12 joints of the hip and leg, defines the leg walking action performance through joint position and key point position ; the motion layer target : controls the whole body root node motion, with linear velocity v , attitude angle rpy and height h as tracking targets, abandoning accurate imitation of leg joint angles.

[0062] In the reinforcement learning training stage, target layering is realized through double reward function design, including:

[0063] Expression reward: , driving accurate imitation of the upper body.

[0064] Motion reward: , guiding the leg to autonomously generate a stable gait.

[0065] After the training based on the above manner is completed, the expression layer target and the motion layer target obtained by the decomposition are analyzed based on the trained strategy model to obtain a control instruction for controlling the robot leg structure. Specifically, first, the desired overall motion is decomposed into two independent and complementary control targets, which are divided into: an expression layer target and a motion layer target, wherein the expression layer target is used to accurately control the positions of 12 joints of the hip and leg to imitate the form and expressiveness of human motion, and the motion layer target is used to control the linear velocity, attitude angle (Roll-Pitch-Yaw) and height of the robot whole body root node to ensure the overall motion stability and balance, then, the human motion data from the motion capture system is adapted to the specific form of the robot through motion reorientation technology to generate a reference motion sequence for the robot to track, including joint position and root node motion trajectory, and finally, based on the deployed trained strategy model, the expression layer and the motion layer target are calculated in real time according to the readings of the current robot body sensors (such as encoders, IMUs, force sensors) to output the desired position, velocity or torque instructions of each joint to drive the hardware to perform the humanoid, stable and terrain-adaptive motion.

[0066] Further, the decomposition processing of the captured motion target is a step before the analysis processing by the strategy model to obtain the control instruction of the target robot. For the captured motion target, which can usually be human motion capture data, the human motion is matched to the corresponding target robot through corresponding analysis processing to control the target robot to reproduce the human motion action. Wherein, in the analysis processing of the captured motion target, it includes: acquiring the captured motion target, and analyzing the motion target by constructing a three-dimensional world coordinate system to obtain a motion capture data sequence in the three-dimensional world coordinate system; based on the motion capture data sequence, calculating the relative rotation angle of the joints of the main body of the motion target in the three-dimensional world coordinate system as a joint position reference sequence, and converting the key point coordinates in the three-dimensional world coordinate system into a local coordinate system coordinate with the main body root node as the origin as a key point position reference sequence; calculating the linear velocity of the main body root node in the three-dimensional world coordinate system as the linear velocity reference sequence of the main body root node, and calculating the attitude angle of the main body root node as the attitude angle reference sequence of the main body root node; acquiring the height of the main body root node in the three-dimensional world coordinate system as the height reference sequence of the main body root node; taking the joint position reference sequence and the key point position reference sequence as the expression layer target, and taking the root node linear velocity reference sequence, the attitude angle reference sequence and the height reference sequence as the motion layer target.

[0067] It should be noted that when the motion target is human motion data, the main body is the human body.

[0068] Exemplarily, when the captured motion target is processed based on the above-mentioned manner to obtain the expression layer target and the motion layer target, it can be divided into the following steps:

[0069] Step one: data preprocessing and coordinate system alignment.

[0070] Firstly, the original motion capture data (i.e. the motion target) is obtained, such as the three-dimensional world coordinates of the marker points (Markers) which can be obtained from optical motion capture systems, IMU suits and other devices, or the three-dimensional poses (positions and rotations) of each skeletal segment of the human body which can be calculated by algorithms. Then, the root node is determined and the corresponding three-dimensional world coordinate system is constructed.

[0071] Step two: motion layer target extraction.

[0072] When the motion layer target is extracted, the linear velocity, attitude angle and height of the root node are extracted, which can be obtained by corresponding calculation and processing in the constructed three-dimensional world coordinate system. For the linear velocity of the root node, the position sequence (x, y, z) of the root node in the world coordinate system can be numerically differentiated (for example, the position difference of the continuous frames is divided by the time interval), to obtain the three-dimensional linear velocity vector v_ref = [vx, vy, vz] of the root node at each time. For the attitude angle of the root node, the Euler angle of the root node in the world coordinate system, i.e. roll, pitch and yaw Ω_ref = [roll, pitch, yaw], can be extracted from the rotation data (usually given in the form of quaternion or rotation matrix) of the root node skeletal segment, to describe the overall inclination and orientation of the human body. For the height of the root node, the z coordinate (vertical height) of the root node in the three-dimensional world coordinate system can be directly taken as the height h_ref.

[0073] Up to now, the motion layer target gm = {v_ref, Ω_ref, h_ref} has been extracted, and the motion layer target describes “how fast the body as a whole should move, in which direction, and in what posture”.

[0074] Step three: expression layer target extraction.

[0075] In the extraction of the expression layer target, joint angles and key point positions are extracted, wherein the key points can be specific positions determined based on the robot or the human body. For the joint angles, in the calculation process, first, a bone chain is defined, a human bone model is constructed according to the motion capture markers, joints corresponding to the robot (such as the hip, knee and ankle) are defined, and then the relative rotation is calculated, that is, for each joint, the rotation of the child bone segment relative to the parent bone segment is calculated. For example, the rotation of the calf bone relative to the thigh bone is calculated, and the knee joint angle is obtained. Finally, the obtained relative rotation (usually converted into Euler angles or exponential mapping form) is combined into a vector, that is, q_ref, representing the target position required by the motor of the joint of the driving robot.

[0076] For the key point positions, in the calculation process, first, key points need to be selected, and bone endpoints that can fully express the motion form, such as six points of two feet, two knees and two hips, are selected, each point has three-dimensional coordinates, and p_ref ∈ R^18, of course, more upper body points can also be included. Then the relative position is calculated, and in order to eliminate the influence of the overall motion, the world coordinates of these key points need to be converted into local coordinates relative to the root node, and then in the calculation process, the calculation formula is p_local = R_root^T * (p_world - root_position), wherein p_world is the world coordinates of the key point, root_position is the world coordinates of the root node, R_root is the rotation matrix of the root node, and p_local is the relative position of the key point in the root node coordinate system calculated. The set of relative positions is the key point position p_ref, which describes "what should the relative attitude of the limbs relative to the torso be regardless of the movement of the body". Thus, the expression layer target ge = {q_ref, p_ref} has been extracted, which describes "what attitude should the joints and limbs assume in order to perform this action".

[0077] In summary, the above embodiment provides a robot leg structure and a control method thereof. The device includes a hip joint module 1, a thigh module 2, a knee joint module 3, a calf module 4 and a foot end module 5. The structure connecting surfaces of the hip joint module 1, the thigh module 2, the knee joint module 3, the calf module 4 and the foot end module 5 are each provided with a flat stop, and the flat stop is connected based on the flat stop. Through the three-degree-of-freedom orthogonal intersection hip joint driving design, the torque motor direct drive knee joint and the modular flat stop connection structure, the problems of motion coupling, transmission precision loss and insufficient assembly precision of the traditional robot lower limbs are effectively solved, and the motion stability is improved, the joint control precision is enhanced, and the dynamic response capability is optimized.

[0078] The robot leg structure and the control method thereof provided by the embodiments of the present application are described in detail above, and the principles and implementation manners of the present application are described by applying specific examples. The above description of the embodiments is only used to help understand the method of the present application and its core idea; meanwhile, for those skilled in the art, the specific implementation manners and application ranges can be changed according to the idea of the present application, and the content of the specification should not be understood as a limitation on the present application. In addition, those skilled in the art can make some improvements and refinements without departing from the principles of the present application, and these improvements and refinements are also regarded as the protection scope of the present application.

Claims

1. A robot leg structure, characterized in that, include: The hip joint module, thigh module, knee joint module, calf module, and foot module are provided with flat plate stops on the structural connection surfaces of the hip joint module, thigh module, knee joint module, calf module, and foot module, and are connected based on the flat plate stops. The height of the flat plate stops is 1.5mm, and the fit tolerance is H6 / g5. The hip joint module includes a first drive unit, a second drive unit, and a third drive unit. The first drive unit controls the roll degree of freedom, the second drive unit controls the pitch degree of freedom, and the third drive unit controls the yaw degree of freedom. The drive axes of the first drive unit, the second drive unit, and the third drive unit are perpendicular to each other and intersect at the same point. The knee joint module is driven by a torque motor, and the rotor of the torque motor is connected to the thigh module, while the stator of the torque motor is connected to the lower leg module. The first drive unit and the second drive unit are planetary reducer collimated drive motors, and the third drive unit is a cycloidal reducer collimated drive motor, and the reduction ratio of the cycloidal reducer is 1:

20. The knee joint module uses a crossed roller bearing as the main support structure. The inner ring of the crossed roller bearing is rigidly connected to the end of the thigh module, and the outer ring of the crossed roller bearing is rigidly fixed to the proximal end of the lower leg module. The knee joint module integrates a non-backlash brake. The first drive unit, the second drive unit, the third drive unit, and the torque motor are fixed based on a spline structure. The module of the spline is 0.5, the pressure angle is 30°, and the fit clearance is less than 0.01mm.

2. The robot leg structure as described in claim 1, characterized in that, The output shaft of the first drive unit is fixedly connected to the pitch frame, and the output shaft of the second drive unit is fixedly connected to the yaw frame.

3. The robot leg structure as described in claim 1, characterized in that, The foot module consists of a foot shell and an elastic foot sole component that are fixedly connected. The foot shell is made of titanium alloy and is integrally formed by 3D printing. The interior of the foot shell has a dot matrix structure. The elastic foot sole component is made of a super-elastic polymer material.

4. The robot leg structure as described in claim 3, characterized in that, The foot module is equipped with a multi-dimensional force sensor and a pressure sensing matrix. The elastic foot planter has multiple built-in miniature deformable cavities, and these deformable cavities are connected to the outside via air tubes.

5. A control method for a robot leg structure, applied to a robot leg structure as described in any one of claims 1 to 4, characterized in that, The method includes: The captured moving target is decomposed into an expression layer target and a motion layer target. The expression layer target is used to constrain the position of the leg joints and the position of key points, and the motion layer target is used to control the linear velocity, attitude angle and height of the robot's whole body root node. The trained policy model is used to analyze and process the expression layer target and the motion layer target, and the control commands for the target robot are output.

6. The control method for the robot leg structure as described in claim 5, characterized in that, The process of decomposing the captured moving target into an expression layer target and a motion layer target includes: The captured moving target is acquired, and a three-dimensional world coordinate system is constructed for the moving target for analysis and processing to obtain a motion capture data sequence in the three-dimensional world coordinate system; Based on the motion capture data sequence, the relative rotation angles of the joints of the moving target's main body in the three-dimensional world coordinate system are calculated as a joint position reference sequence, and the key point coordinates in the three-dimensional world coordinate system are converted into local coordinate system coordinates with the root node of the main body as the origin, as a key point position reference sequence. The linear velocity of the main body root node is calculated in the three-dimensional world coordinate system as the linear velocity reference sequence of the main body root node, and the attitude angle of the main body root node is calculated as the attitude angle reference sequence of the main body root node. The height of the main root node in the three-dimensional world coordinate system is obtained and used as the height reference sequence of the main root node. The joint position reference sequence and the key point position reference sequence are used as the expression layer targets, and the root node linear velocity reference sequence, the attitude angle reference sequence, and the height reference sequence are used as the motion layer targets.

Images