Humanoid robot

Abstract

The utility model discloses a humanoid robot, the trunk subassembly, femur subassembly and shank bone subassembly of this humanoid robot are provided with multiple inertia measurement units, the footboard assembly includes first, second, third, fourth ground contact subassembly, and it is connected in series through the pivot assembly in proper order, the first ground contact subassembly is located the forward direction of the trunk subassembly, the fourth ground contact subassembly is located the rear direction of the trunk subassembly, the fourth ground contact subassembly is connected through a pivot assembly on the downside of the shank bone subassembly, the footboard assembly is installed with pivot assembly and pressure sensing unit, through the detection of multiple inertia measurement units and multiple pressure sensing components, the humanoid robot can carry out ZMP zero moment point gait control, MPC model predictive control analysis, to reach the purpose of robot autonomous detection, barycentre balance.

Description

Technical Field

[0001] This utility model relates to a robot, and more particularly to a humanoid robot with a special footplate structure design that can achieve artificial intelligence control or autonomous control, and can also autonomously balance its movement. Background Technology

[0002] Humanoid robots possess bipedal locomotion and bi-arm operation capabilities similar to humans, as well as multimodal perception functions including vision, hearing, force, and touch. They can move and perform tasks in typical indoor environments such as flat ground and stairs. However, due to the high center of gravity of humanoid robots, their safe and stable movement requires extremely high control standards, making it difficult to achieve high-speed movement. Especially in environments with inclines, declines, stair climbing, stair descending, obstacles, and uneven surfaces, the ability of humanoid robots to navigate, maintain balance, and prevent falls on complex terrain remains a significant challenge.

[0003] For example, invention patent CN201580020076.3 discloses an omnidirectional wheeled humanoid robot, which has the advantage of rapid movement, but cannot smoothly cross obstacles. Utility model patent CN201720886723.5 discloses a device for switching the movement modes of a humanoid robot. A rotating power component drives rollers to rotate relative to a base plate, allowing the robot to switch between low-speed and high-speed modes, but it cannot perform legged movement, and the humanoid robot will lose balance on uneven terrain. Invention patent CN201510443721.4 discloses a moving humanoid robot and its working method. This robot achieves wheeled walking by setting drive wheels and driven wheels on the outside of its legs. The drive wheels are driven by a motor. To a certain extent, this solution improves the movement speed and balance of the humanoid robot, but it requires an additional drive motor to achieve highly complex movements, thus increasing the robot's complexity and weight, and reducing control stability.

[0004] Therefore, how to fix the center of gravity of the humanoid robot, maintain its balance, improve its terrain adaptability, smoothly maintain the stability of the center of gravity on slopes and uneven ground, achieve consistent perception and motion control in complex environments, overcome the lack of steering flexibility, and solve the posture control problem in dynamic forward movement or rotation states is a major challenge that technical workers in this field need to face. Utility Model Content

[0005] (a) Technical problems to be solved

[0006] The main purpose of this invention is to enable humanoid robots to have dynamic center of gravity control, thereby allowing them to adapt to complex terrains such as uphill, downhill, undulating, and uneven terrain.

[0007] Another objective of this invention is to address the problem of excessive energy consumption during walking caused by the rigid foot structure of bipedal robots, which lacks energy storage and impact absorption mechanisms.

[0008] Another objective of this invention is to enable humanoid robots to maintain a fixed center of gravity and balance when moving and rotating, so as not to lose stability and fall.

[0009] (II) Technical Solution

[0010] To address the aforementioned and other problems, this utility model provides a humanoid robot, comprising a torso assembly, two femur assemblies, two tibialis assemblies, two footplate assemblies, and multiple inertial measurement units. The femur assemblies are connected to the torso assembly at the top and pivotally connected to the tibialis assemblies at the bottom. The tibialis assemblies are connected to the footplate assemblies at the bottom. The robot is characterized in that at least one inertial measurement unit is provided on each of the torso assembly, femur assemblies, and tibialis assemblies to detect the movement and rotational inertia of various parts of the humanoid robot. The footplate assemblies include a first ground contact assembly, a second ground contact assembly, a third ground contact assembly, and a fourth ground contact assembly. The robot comprises a torso assembly, multiple pivot assemblies, and multiple pressure sensing units. The first, second, third, and fourth ground contact assemblies are positioned facing the ground or near the periphery of the ground and are sequentially connected in series via the pivot assemblies. The first ground contact assembly is located in the forward direction of the torso assembly, and the fourth ground contact assembly is located in the rearward direction of the torso assembly. The fourth ground contact assembly is connected to the lower side of the shinbone assembly via a pivot assemblies. The humanoid robot performs ZMP zero-moment point gait control and analysis through the inertial measurement unit and performs center of gravity balance feedback correction through the pressure sensing units.

[0011] In the humanoid robot described above, the femur assembly has two inertial measurement units, which are respectively located at both ends of the femur assembly along the axial direction; or, the shinbone assembly has two inertial measurement units, which are respectively located at both ends of the shinbone assembly along the axial direction.

[0012] In the humanoid robot described above, at least one inertial measurement unit is provided on the first, second, third, or fourth ground contact assembly to detect its movement and rotational inertia; in a further embodiment, the inertial measurement unit of the first, second, third, or fourth ground contact assembly is provided on the pivot assembly or facing the ground.

[0013] In the humanoid robot described above, at least one buffer rod extends from the third ground contact component and is connected to the second ground contact component.

[0014] The humanoid robot described above, wherein the second grounding component extends in a rearward direction and is connected to the fourth grounding component via a pivot component.

[0015] (III) Beneficial Effects

[0016] Compared with the prior art, the humanoid robot of this utility model has the following beneficial effects:

[0017] 1. The specially designed foot plate assembly can adapt to complex terrains such as uphill, downhill, undulating, and uneven surfaces.

[0018] 2. To address the problem of excessive energy consumption during walking caused by the lack of energy storage and impact absorption mechanisms in the rigid foot structure of bipedal robots. Attached Figure Description

[0019] Figures 1-2 The diagram shown is a structural schematic of each part of the humanoid robot described in this utility model.

[0020] Figure 3 The diagram shown is an exploded view of the foot plate assembly of the first embodiment.

[0021] Figures 4-7 The illustration shows the operation of the foot plate assembly of the first embodiment under different conditions.

[0022] Figure 8 The diagram shown is an exploded view of the foot plate assembly of the second embodiment.

[0023] Figures 9-10 The illustration shows the operation of the foot plate assembly under different conditions in the second embodiment.

[0024] Figure 11 The diagram shown is a flowchart of the posture control method for the humanoid robot of this utility model.

[0025] Figure 12 The diagram shown is a flowchart of the ZMP zero-moment point gait analysis method of this utility model.

[0026] Figures 13-14 The diagram shows the variation of the support polygon in the gait analysis of the ZMP zero moment point.

[0027] Figure 15 The diagram shows the humanoid robot's foot assembly in a state of going uphill.

[0028] Figure 16 The diagram shows the foot assembly of the humanoid robot in a downhill situation.

[0029] Figure 17 The diagram shows the humanoid robot's foot assembly accelerating.

[0030] Figure 18 The diagram shows the humanoid robot's foot assembly during deceleration.

[0031] Figure 19 The illustration shows the humanoid robot adjusting its center of gravity.

[0032] Figure 20 The illustration shows the humanoid robot tilting to the right.

[0033] Figure 21 The diagram shows the method flowchart of the MPC predictive control algorithm in step A6.

[0034] Figure 22 The diagram shows the flowchart of the method for calculating the stable equilibrium of joint angle θ using the Lyapunov stability function in step A7.

[0035] Explanation of reference numerals in the attached figures:

[0036] 10-Humanoid robot; 11-Thigh bone assembly; 12-Lower leg bone assembly; 13-Tortoise assembly; 14-Foot plate assembly; 141-First ground contact assembly; 142-Second ground contact assembly; 143-Third ground contact assembly; 144-Fourth ground contact assembly; 145-Pivot assembly; 146-Pressure sensing unit; 147-Buffer rod; 15-Inertial measurement unit; FD-Forward direction; BD-Rearward direction; θ-Joint angle; 91-Support polygon; 92-Zero torque point. Detailed Implementation

[0037] With the advancement of AI, the applications of robots are increasing and gradually expanding into all aspects of human life. Influenced by factors related to human beings, the development and application of humanoid or humanoid robots are gradually becoming a key research focus in the field of robotics. Please refer to [link / reference]. Figure 1 , Figure 2 , Figures 1-2The diagram illustrates the structure of various parts of the humanoid robot described in this invention. As shown, a humanoid robot 10 includes a torso assembly 13, two femur assemblies 11, two shinbone assemblies 12, two footplate assemblies 14, and multiple inertial measurement units 15. The femur assemblies 11 are connected to the torso assembly 13 at the top and pivotally connected to the shinbone assemblies 12 at the bottom. The shinbone assemblies 12 are connected to the footplate assemblies 14 at the bottom. The femur assemblies 11 and shinbone assemblies 12 are connected at a joint angle θ. Multiple inertial measurement units 15 are provided on the torso assembly 13, femur assemblies 11, and shinbone assemblies 12 to detect the movement and rotational inertia of various parts of the humanoid robot 10. The inertial measurement unit 15 (IMU) can be used to measure the three-axis attitude angles (or angular rates) and acceleration of an object. Generally, an IMU (Inertial Measurement Unit) 15 contains three-axis gyroscopes and three-axis accelerometers (x, y, z), enabling it to measure the angular velocity and acceleration of an object in three-dimensional space, and thereby calculate the object's speed, impulse, and other dynamic changes. This inertial measurement unit 15 is mostly used in devices requiring motion control, such as automobiles and robots, and can also be used in applications requiring precise displacement calculations based on attitude, such as inertial navigation equipment for submarines, aircraft, missiles, and spacecraft. Figure 1 , Figure 2 In one embodiment, the femoral bone assembly 11 has one inertial measurement unit 15, which is located at the lower end of the femoral bone assembly 11. The shinbone assembly 12 has two inertial measurement units 15, which are located at the upper and lower ends of the shinbone assembly 12, respectively. In other embodiments, the more inertial measurement units 15, the better. More inertial measurement units 15 can provide more accurate dynamic detection or eliminate single-point blind spots, thereby improving the attitude estimation accuracy of the humanoid robot 10. In addition, setting more inertial measurement units 15 can also be used to detect the slight bending or rotation of the femoral bone assembly 11 and the shinbone assembly 12, so as to accurately calculate the mechanical stress state between the components and to serve as a redundancy compensation design. In other embodiments, the number of inertial measurement units 15 on the femoral bone assembly 11 and the shinbone assembly 12 can also be one or more, depending on cost considerations or functional considerations.

[0038] First Embodiment

[0039] Please also refer to Figure 3 , Figures 4-7 , Figure 3 The diagram shown is an exploded view of the foot plate assembly according to the first embodiment. Figures 4-7 The illustration shows the operation of the footplate assembly of the first embodiment under different conditions. Figure 3 As shown, the footplate assembly 14 includes a first ground contact assembly 141, a second ground contact assembly 142, a third ground contact assembly 143, a fourth ground contact assembly 144, multiple pivot assemblies 145, and multiple pressure sensing units 146 (also labeled as...). Figure 1 , Figure 2 The first, second, third, and fourth ground contact components 144 are each provided with at least one pressure sensing unit 146 facing the ground or near the periphery of the ground. The first, second, third, and fourth ground contact components 144 are sequentially connected in series via the pivot assembly 145. The first ground contact component 141 is located in the forward direction FD of the torso component 13, and the fourth ground contact component 144 is located in the rearward direction BD of the torso component 13. The fourth ground contact component 144 is connected to the lower side of the shinbone component 12 via a pivot assembly 145 (see figure). Figure 1 , Figure 2 The pressure sensing unit 146 is a piezoelectric material that, when subjected to mechanical stress or pressure, can generate charge separation (i.e., generate voltage output), or can deform when voltage is applied. This unique bidirectional effect, known as the "piezoelectric effect," makes this type of material widely used in sensors, transducers, actuators, and many other fields. Generally, piezoelectric materials have two types: direct piezoelectric effect and inverse piezoelectric effect. In this embodiment, the direct piezoelectric effect is used. That is, when mechanical stress is applied to the pressure sensing unit 146, an asymmetrical distribution of charge is generated inside the material, thereby forming a voltage difference on the material surface. Then, by observing the magnitude of this voltage difference, the magnitude of the ground reaction force when different parts of the first, second, third, and fourth ground contact components 144 touch the ground can be determined. In the humanoid robot 10 of this invention, the multiple pressure sensing units 146 provided in the first, second, third, and fourth ground contact components 144 can achieve the effect of multi-point sensing and flexible detection. Furthermore, to better understand the subtle terrain changes at each part of the footplate assembly 14, at least one inertial measurement unit 15 can be provided on the first ground contact assembly 141, the second ground contact assembly 142, the third ground contact assembly 143, and the fourth ground contact assembly 144 to detect the movement and rotation vectors of the first, second, third, and fourth ground contact assemblies 144. In a preferred embodiment, the inertial measurement unit 15 of the first ground contact assembly 141, the second ground contact assembly 142, the third ground contact assembly 143, or the fourth ground contact assembly 144 can be located on the pivot assembly 145 or facing the ground. Figure 4 As shown, the first, second, third, and fourth ground-contact components 144 are connected in series to form the footplate assembly 14. Figure 5As shown, through the control of the left pivot assembly 145 of the third ground contact assembly 143 (see comparison) Figure 3 This allows the third ground contact component 143 to rotate counterclockwise. For example... Figure 6 As shown, by controlling the pivot assembly 145 on the left side of the first ground contact assembly 141, the first ground contact assembly 141 can be rotated counterclockwise or clockwise (e.g., ...). Figure 7 (As shown).

[0040] Second Embodiment

[0041] Please also refer to Figure 8 , Figures 9-10 , Figure 8 The diagram shown is an exploded view of the foot plate assembly according to the second embodiment. Figures 9-10 The illustration shows the operation of the footplate assembly of the second embodiment under different conditions. Figure 8 As shown, the footplate assembly 14 of the second embodiment is also composed of the first ground contact assembly 141, the second ground contact assembly 142, the third ground contact assembly 143, and the fourth ground contact assembly 144. The fourth ground contact assembly 144 is disposed below the shinbone assembly 12, and is connected to the third ground contact assembly 143 via a pivot assembly 145. Two retractable and extendable buffer rods 147 extend from the upper side of the third ground contact assembly 143, and the two buffer rods 147 are connected to the upper second ground contact assembly 142. The second ground contact assembly 142 extends in a rearward direction BD and is connected to the fourth ground contact assembly 144 via a pivot assembly 145. Please refer to [further details omitted]. Figure 9 When the first, second, third, and fourth ground contact components 144 are connected and assembled, the two buffer rods 147 can be extended to move the second ground contact component 142 and the third ground contact component 143 away from each other; for example... Figure 10 As shown, when the two buffer bars 147 are shortened, the second ground contact assembly 142 and the third ground contact assembly 143 can move closer to each other. That is, when the foot plate assembly 14 is subjected to the ground reaction force, the foot plate assembly 14 can alleviate the impact force of the ground through the two buffer bars 147; or, the two buffer bars 147 can be actively pushed or pulled (i.e., actively extended or shortened) to control the force of the foot plate assembly 14 bouncing off the ground.

[0042] In this way, through the above structural design, the humanoid robot 10 of this invention can achieve multi-dimensional dynamic balance control and terrain adaptability through the integrated design of a composite sensing system and a bionic foot structure. Specific benefits include: the ability to adapt to complex terrain. By using IMU inertial measurement units 15 installed on the torso assembly 13, femur assembly 11, and shinbone assembly 12, combined with multiple pressure sensing units 146 at different locations on the foot assembly 14, the robot can detect the inertia of each part and the distribution of ground contact pressure in real time. Then, through ZMP zero-torque point calculation and feedback correction, the robot can dynamically adjust its center of gravity position and actively adapt to uphill, downhill, uneven, and irregular terrain. Furthermore, this structure boasts high stability and low energy consumption. Through the series design of multiple ground-contact components in the foot plate assembly 14, combined with the structural design of the pivot assembly 145 and the buffer rod 147, the rigid foot can be transformed into a flexible multi-segment ground-contact unit. Thus, the foot plate assembly 14 not only absorbs walking impact and stores kinetic energy to reduce energy consumption, but also accurately interprets ground reaction force through distributed pressure sensing. Combined with the bidirectional inertial monitoring of the IMU inertial measurement units 15 at both ends of the femur assembly 11 and shinbone assembly 12, this significantly improves the motion stability of the humanoid robot 10, preventing instability and falls, and achieving gait analysis and control. Moreover, this structure can assist in autonomous intelligent decision-making. That is, through the inertial measurement unit 15 and the pressure sensing unit 146, the robot's motion data and pressure information can be integrated for collaborative analysis, allowing the robot to autonomously adjust its gait and posture without human intervention, achieving flexible turning and dynamic balance, enhancing the synchronization of environmental perception and motion control, and realizing human-like environmental adaptive intelligence.

[0043] Method Implementation Examples

[0044] Please see Figure 11 , Figure 11 The diagram shown is a flowchart of the posture control method for the humanoid robot of this utility model. Figure 11As shown, the humanoid robot 10 described in this utility model is first provided (step A1). Then, the motion state of each part of the humanoid robot 10 is detected in real time through the inertial measurement unit 15 already installed on the humanoid robot 10 (step A2). Then, the detection signals of multiple inertial measurement units 15 and pressure sensing units 146 are obtained to distinguish the type of imbalance of the humanoid robot 10 (step A3). Generally speaking, the road conditions that the humanoid robot 10 may face include uphill, downhill, acceleration or deceleration when running, and falling over (losing its center of gravity) when standing on both feet due to external force. From a dynamic perspective, the dynamic problems faced by the humanoid robot 10 involve changes in position, acceleration, and angular acceleration. Therefore, if the humanoid robot wants to autonomously handle the autonomous balance under external force or predict the action in uneven situations, so as to achieve the purpose of limb correction and center of gravity adjustment, it is necessary to perform a preliminary classification of the balance issues of the humanoid robot 10. In this way, the applicable algorithm or mathematical function can be quickly identified in subsequent analysis and calculation. For example, if the pressure detected by the pressure sensing unit 146 in the forward direction FD of the foot assembly 14 is less than that in the backward direction BD, step A3 can determine that the humanoid robot 10 may be traveling downhill; if the pressure detected by the pressure sensing unit 146 in the forward direction FD of the foot assembly 14 is greater than that in the backward direction BD, then step A3 can determine that the humanoid robot 10 may be traveling uphill. Of course, actual ground conditions may be more complex, and uphill and downhill are only preliminary classifications, which still need to be supplemented by further analysis. Next, based on the ZMP zero moment point 92 gait control, an optimal balance strategy is calculated and analyzed (step A4). Here, in order to perform ZMP zero moment point gait analysis on the bipedal robot, the definitions are "Supporting Polygon" and "ZMP Zero Moment Point". ZMP, short for Zero Moment Point, also known as the Zero Moment Point or Center of Pressure, refers to the point on the ground where the resultant force of all horizontal momentum of the humanoid robot 10 is zero. In Miomir Vukobratovic's equilibrium theory, as long as there is a ZMP zero moment point within the "support polygon" of the bipedal robot's feet—or, in other words, the intersection of the extension of the inertial force with the ground lies within the "support polygon"—the multi-legged system will not tip over. Therefore, whether stationary or walking, if the ZMP zero moment point is outside the range of the "support polygon," the humanoid robot 10 will tip over; if the ZMP zero moment point falls within the range of the support polygon, the humanoid robot 10 can maintain balance.This also means that in the gait control of the humanoid robot 10, real-time calculation or determination of the position of the zero-moment point is crucial when it undergoes dynamic changes such as movement or rotation, as this directly relates to the robot's balance and stability. Therefore, ensuring that the ZMP zero-moment point of the humanoid robot 10 is always located within the supporting polygon can maintain the balance and stability of the humanoid robot 10, whether it is on a slope or running. Please refer to [link / reference]. Figure 13 , Figure 14 , Figures 13-14 The diagram illustrates the variation of the support polygon in ZMP zero-moment point gait analysis. Here, the support polygon 91 refers to the area enclosed by all the feet of the multi-legged robot in contact with the ground. Taking the bipedal system of humanoid robot 10 as an example, it can be divided into two states: single-leg support and dual-leg support. Figure 13 As shown, when the humanoid robot 10 is supported on the ground by its two feet, its "support polygon 91" is the convex polygon formed by the area of ​​its feet in contact with the ground. When both feet are on the ground, the movement of the humanoid robot 10's limbs, the swinging of its hands, and the movement and swaying of its torso component 13 will all cause a change in the position of the zero-moment point 92 of the humanoid robot 10. As long as the range of movement of the zero-moment point 92 does not exceed the support polygon 91, the humanoid robot 10 can maintain its balance and stability and will not fall or overturn. If the movement of the limbs, the swinging of the hands, and the movement and swaying of the torso component 13 are too great, causing the range of movement of the zero-moment point 92 to exceed the support polygon 91, then the humanoid robot 10 will be unable to maintain its balance and stability. For example... Figure 14 As shown, when one leg is supporting the ground, the "support polygon 91" represents the area of ​​the sole of the supporting leg touching the ground. In a walking or running scenario, the zero-moment point 92 sways significantly. Therefore, the other leg, which is suspended in the air, needs to be analyzed and calculated to ensure that the zero-moment point 92 is back within the support polygon 91. Thus, the ZMP (Zero Moment Point) algorithm is applied to the balance control calculation of the humanoid robot 10. Its core idea is to adjust the zero-moment point 92 of the humanoid robot 10 to keep it within the support polygon 91, ensuring the robot does not fall over. Then, by monitoring and adjusting the position of the ZMP, the humanoid robot 10 achieves both static and dynamic balance.

[0045] Please see Figure 12 , Figure 12 The diagram shown is a flowchart of the ZMP zero-moment point gait analysis method of this utility model. Figure 12As shown, in order to achieve the ZMP gait analysis calculation in step A4, A4 further includes the following sub-steps: analyzing the center of mass position and trajectory change of the humanoid robot 10 through multiple inertial measurement units 15 (step A41), and collecting dynamic information such as velocity, acceleration, angular velocity, and angular acceleration of the humanoid robot 10 in real time through the IMU inertial measurement units 15 installed on the torso component 13, femur component 11, and shinbone component 12; in a preferred embodiment, a filtering algorithm (such as a Kalman filter) can even be introduced to process the data of the IMU inertial measurement units 15 to improve the accuracy of attitude estimation; based on the collected dynamic data, combined with the 3D structure and geometric space modeling distribution of the humanoid robot 10, the current center of mass position of the humanoid robot 10 and its trajectory (i.e., acceleration) over time can be calculated. Next, the momentum change of the humanoid robot 10 is obtained through the inertial measurement unit 15 on the torso component 13 (step A42). The velocity and acceleration data obtained by the IMU inertial measurement units 15 on the femur component 11 and tibia component 12 are analyzed. Combined with the 3D spatial structure of the humanoid robot and its current velocity, the total momentum change of the humanoid robot 10 can be calculated. This dynamic information is extremely important for analyzing the motion state of the humanoid robot 10, especially during rapid movement or changes in direction. Then, the supporting polygon 91 of the humanoid robot 10 is confirmed through multiple pressure sensing units 146 (step A43). Please also refer to... Figure 13 and Figure 14 The distribution of ground reaction force is detected by multiple pressure sensing units 146 installed on the foot plate assembly 14 of the humanoid robot 10. Combined with multiple IMU inertial measurement units 15 on the torso assembly 13, femur assembly 11, and shinbone assembly 12, the range of the supporting polygon 91 of the humanoid robot 10 in real-time can be analyzed and determined based on the pressure values ​​recorded by each sensor. Then, according to the Newton-Euler equations, the relative positional relationship between the ZMP zero moment point 92 of the humanoid robot 10 and the supporting polygon 91 of the humanoid robot 10 is calculated and analyzed (step A44). The specific formula of the Newton-Euler equations is:

[0046]

[0047] Where Xzmp is the x-coordinate of the zero-moment point 92 of ZMP, X CoM It is the x-coordinate of the center of mass of the humanoid robot, a. z It is acceleration, g is the acceleration due to gravity, X fThe x-coordinate is the point of contact between the foot and the ground. CoM refers to the Center of Mass. Similarly, the formula for the y-coordinate is the same as that for the x-coordinate; simply replace X with Y. Thus, by using the CoM center of mass coordinate information obtained from the IMU inertial measurement unit 15 and the pressure sensing unit 146, along with the coordinate information of the foot assembly 14, and combining this with calculations based on the Newton-Euler equations, the spatial positional relationship between the ZMP zero-moment point 92 and the supporting polygon 91 of the humanoid robot 10 can be determined. Subsequently, if the position of the ZMP zero torque point 92 exceeds the range of the support polygon 91 of the humanoid robot 10, the movement vector that the torso component 13, femur component 11, or shinbone component 12 needs to balance can be determined and calculated (step A45). Then, the angle, position, and other step parameters of the torso component 13, femur component 11, and shinbone component 12 can be adjusted. That is, the torso component 13, femur component 11, or shinbone component 12 can form a new support polygon 91 by changing its action at the next time point, so as to achieve the posture correction of the humanoid robot 10 and achieve the purpose of stable balance.

[0048] Once the ZMP analysis in step A4 yields a result, the limbs of the humanoid robot 10 can be driven to extend, bend, or adjust its center of gravity (step A5). Below are some scenarios that the humanoid robot 10 might encounter in practice, illustrating the methods or steps in step A5 for the humanoid robot 10 to extend and bend its limbs to adjust its center of gravity and achieve body balance. Please refer to... Figure 15 , Figure 15 The illustration shows the foot assembly of the humanoid robot in a state of going uphill; as shown. Figure 15 As shown, when the humanoid robot 10 moves uphill, step A5 first drives the fourth ground contact component 144 and the third ground contact component 143 to land simultaneously, and then sequentially drives the second and first ground contact components 141 to land. The landing sequence and take-off sequence of each ground contact component of the foot assembly 14 can be controlled through the pivot assembly 145 (see also...). Figure 3 This is used to mitigate ground impact and conserve energy. Please refer to [further details]. Figure 16 , Figure 16 The illustration shows the foot assembly of the humanoid robot in a downhill situation; as shown Figure 16 As shown, when the humanoid robot 10 moves downhill, step A5 sequentially drives the fourth, third, second, and first ground-contact components 141 to touch the ground. It should be specifically noted that the contact sequence of each ground-contact component of the foot assembly 14 is the same when the humanoid robot 10 moves downhill and on flat ground; thereby, the humanoid robot 10 can also mitigate ground impact and save energy consumption when moving on flat ground or downhill.

[0049] Please see Figure 17 , Figure 17 The illustration shows the humanoid robot's foot assembly accelerating. Figure 17 As shown, when the two foot plate assemblies 14 alternately touch the ground, step A5 drives the foot plate assembly 14 in contact with the ground, causing the first ground contact component 141 to momentarily push against the ground, thereby accelerating forward; that is, the pushing force of the first ground contact component 141 can increase the acceleration value compared to simply using the thigh bone assembly 11 and the shin bone assembly 12. Please refer to... Figure 18 , Figure 18 The diagram illustrates the deceleration process of the humanoid robot's foot assembly. Figure 18 As shown, when the two footplate assemblies 14 alternately touch the ground (i.e., during running or brisk walking), step A5 drives the suspended footplate assembly 14, causing its fourth ground-contact component 144 to momentarily push against the ground to exert greater instantaneous resistance and achieve a deceleration effect. Please refer to... Figure 19 , Figure 19 The illustration shows the humanoid robot adjusting its center of gravity; as shown Figure 19 As shown, when the humanoid robot 10 stands on two legs, but momentarily loses its center of gravity due to external force, step A44 calculates and analyzes that the ZMP zero-moment point 92 moves towards the outer periphery of the supporting polygon 91, and then exceeds the range of the supporting polygon 91, thus beginning to tip over; at this time, step A5 can drive the thigh bone assembly 11 and shin bone assembly 12 of the humanoid robot 10 to bend (producing an effect similar to a human squatting down), thereby lowering the center of gravity of the entire humanoid robot 10. Please refer to Figure 20 , Figure 20 The illustration shows the humanoid robot tilting to its right; as shown Figure 20As shown, when the humanoid robot 10 tilts to the right (based on the field of vision of the humanoid robot 10), step A44 calculates and analyzes that the ZMP zero moment point 92 moves toward the right side of the torso assembly 13. Step A5 then drives the right foot plate assembly 14 and the right femur assembly 11 and shinbone assembly 12 of the humanoid robot 10 to rotate clockwise, so that the supporting polygon 91 of the humanoid robot 10 covers the ZMP zero moment point 92. That is to say, the zero moment point 92 can be placed inside the supporting polygon 91 after the right foot plate assembly 14 rotates. In this way, the humanoid robot 10 will no longer continue to tilt to the right due to the rotation of the right foot plate assembly 14. Similarly, when step A44 calculates and analyzes that the ZMP zero torque point 92 moves toward the left side of the torso assembly 13, step A5 drives the left foot plate assembly 14 of the humanoid robot 10 to rotate counterclockwise (not shown); this changes the support polygon 91 of the humanoid robot 10, thereby preventing the humanoid robot from tilting to the left.

[0050] Next, the humanoid robot 10 may face complex, varied, and unpredictable road conditions in real-life scenarios. Therefore, it also needs to be able to predict gait over a period of time and provide feedback corrections to the gait parameters of the humanoid robot 10 within this cycle to achieve a rolling optimization effect. Therefore, step A6 (see...) can also be added. Figure 11 The system uses information collected by multiple inertial measurement units 15 and pressure sensing units 146 to iterate and optimize the gait using the MPC (Model Predictive Control) algorithm. Here, MPC is a programmed control method based on a dynamic model of the process. Under specific constraints, it can optimize the step distance for each time period. That is, the MPC algorithm captures the robot's dynamics at past time points and then analyzes and calculates the possible events for the next cycle (for humanoid robot 10, this could be uneven terrain, swaying acceleration during running, or sudden external forces, etc.), thereby planning the optimal response for humanoid robot 10 (e.g., deceleration, lowering the center of gravity, changing individual step distances, etc.). Please refer to... Figure 21 , Figure 21 The diagram shown is a flowchart of the MPC predictive control algorithm in step A6, as follows: Figure 21As shown, the sub-steps of step A6 include: obtaining the attitude information and ground reaction force information of the humanoid robot 10 through the inertial measurement units 15 on the torso assembly 13, the femur assembly 11 and the shinbone assembly 12 (step A61); synchronously collecting acceleration and angular velocity data of the x, y and z axes from the IMU inertial measurement units 15 installed at the joints of the torso assembly 13, the femur assembly 11 and the shinbone assembly 12 at a frequency of approximately 1 kHz; and collecting the ground reaction force at a frequency of 500 Hz from the pressure sensing unit 146 of the foot assembly 14, thereby calculating the roll angle, pitch angle, yaw angle and other information of the humanoid robot 10, and calculating the components of the ground reaction force, and even deducing the joint angle θ, angular velocity and acceleration through the kinematic model. Next, based on the obtained information, a dynamic model of the humanoid robot 10 is established, and the current position of the center of mass and its velocity are calculated accordingly (step A62). In a preferred embodiment, the Lagrangian equation can be used to analyze the mass, momentum, or angular momentum of the two-legged platform assembly 14 to obtain the kinetic energy and potential energy of the two legs. This method treats the legs of the humanoid robot 10 as swinging supports, so its dynamic model can also be called the "swinging leg model". The Lagrangian equation is as follows:

[0051]

[0052] Where, q sLet Ds be the joint angle vector, Cs be the inertia matrix, Gs be the Coriolis matrix, Bs be the joint torque, and u be the torque of each joint. Here, the Coriolis force Cs can be ignored in the context of bipedal robots. Thus, by analyzing the dynamic model of the bipedal robot's rigid body using Lagrange equations, parameters such as the mass, momentum, angular momentum, joint friction, and external forces of the leg bones are incorporated, thereby deriving the position and velocity of the center of mass within each control cycle (approximately 10 milliseconds). Subsequently, the center-of-mass trajectory of the humanoid robot 10 within the next N steps is predicted using the dynamic model (step A63). Within the MPC framework, the dynamic model can be linearized into a state-space model and discretized into a time step Δt = 10 ms. The prediction time domain is set to N = 50 steps (total time 0.5 seconds). The Lagrange equation is solved in each control cycle to minimize the center-of-mass trajectory deviation, ZMP error, and joint torque, thereby generating a sequence of predicted matrix matrices for the center-of-mass displacement, velocity, and acceleration for the next N steps. Then, based on the predicted center-of-mass trajectory, a set of gait matrix sequences for the next N steps is planned, and the target position and attitude of the humanoid robot 10 for each step are determined (step A64). Based on the planned center-of-mass trajectory, foot trajectory (including swing and support polygons), center-of-mass height change, and upper body attitude angle for the next N steps of the humanoid robot 10, the target position and attitude angle (roll, pitch, yaw) for each step are set, and joint reference trajectories (angular position, angular velocity, angular acceleration) are generated. Then, the motion state of the humanoid robot is monitored in real time, and its actual motion state is compared with the predicted center of mass trajectory (step A65). Here, real-time state monitoring and error analysis are performed. In each control cycle, the measured center of mass position, velocity, and ZMP position are compared with the predicted matrix sequence values, and the error vector is calculated. Specifically, the position error threshold is set to 0.01m and the velocity error threshold is set to 0.05m / s. When the error threshold at any time point exceeds this range, the gait correction mechanism is triggered. When the error between the actual motion state and the predicted center of mass trajectory exceeds the predetermined threshold, the target position in the subsequent gait matrix sequence set and the posture information of the humanoid robot 10 are fed back (step A66). That is, when the error exceeds the limit, the feedback result is introduced into the Lagrange equation, and the real-time error is input into the dynamic matrix for iteration, thereby adjusting the prediction of the center of mass trajectory for the next N steps, so as to achieve real-time rolling correction and optimization of the posture information of the humanoid robot 10 (step A67), forming a closed-loop control. In this way, by using the MPC model predictive control algorithm in step A6 and introducing the Lagrange method for data matrix calculation and iteration, the gait of the bipedal robot can be effectively fed back in real time, adjusted and optimized immediately, ensuring that it can maintain good balance and stability in various complex environments.Therefore, through the analysis and calculation of the MPC model predictive control algorithm in step A6, the gait of the bipedal robot can be effectively iterated, fed back, and adjusted in real time, thereby achieving the effect of predicting, planning, and optimizing the gait. This ensures that the humanoid robot 10 can maintain good balance and stability in various complex environments. Furthermore, it enables the humanoid robot 10 to achieve autonomous balance and posture adjustment under various motion conditions such as starting from a standstill, decelerating while moving, going uphill, going downhill (with inclines and declines of uphill and downhill degrees of ±25 degrees), and uneven surfaces. This allows the humanoid robot 10 to autonomously adjust its dynamic center of gravity within a response time of less than 50 milliseconds.

[0053] like Figure 11 As shown, in addition to the MPC model predictive control optimization in step A6, this invention can also detect the gait stability of the humanoid robot 10 through the Lyapunov stability function algorithm. That is, through the information collected by multiple inertial measurement units 15 and pressure sensing units 146, the Lyapunov stability function is used to perform stable balance calculation of the joint angle θ and dynamic control of the humanoid robot 10 (step A7). Please also refer to... Figure 22 , Figure 22 The diagram illustrates the method for calculating the stable equilibrium of joint angle θ using the Lyapunov stability function in step A7; as shown below. Figure 22 As shown, based on the motion state of the humanoid robot 10 obtained by the inertial measurement unit 15, a dynamic model of the Lyapunov function is established (step A71). First, the gait information such as attitude, velocity, and acceleration of the humanoid robot 10 is obtained through the IMU inertial measurement unit 15 installed on the thigh bone assembly 11 and the shin bone assembly 12. The joint angle θ between the thigh bone assembly 11 and the shin bone assembly 12 is calculated. The joint angle θ, angular velocity, angular acceleration, external force, and torque of the humanoid robot 10 are incorporated to construct a dynamic model of the Lyapunov stability function. The dynamic model of the Lyapunov stability function is as follows:

[0054]

[0055] Where q is the joint angle vector, τ is the joint torque, and M, C, and g are the inertia matrix, Coriolis force term, and gravity term, respectively. In this application of the humanoid robot 10, the Coriolis force can be ignored. Here, Lyapunov stability can be used to describe the stable state of a dynamic system. If the dynamic system is given an initial condition, and its trajectory can be maintained near the initial condition for a continuous period of time, then the system is "Lyapunov stable" at that initial condition. Next, the angular velocity and angular acceleration of the joint angle θ are converted from the starting acceleration or deceleration acceleration of the humanoid robot 10, and the function boundary conditions of the humanoid robot 10 are established accordingly (step A72). That is, after establishing the Lyapunov stability equation, the angular velocity and angular acceleration of the joint angle θ are converted from the starting acceleration or deceleration acceleration of the humanoid robot 10, and the function boundary conditions are established according to the physical structure and motion constraints of the humanoid robot 10. These boundary conditions include the range of joint angles θ, and the upper and lower limits of angular velocity and angular acceleration. For example, the angles of certain joints may only be able to move within a specific range; exceeding this range may cause damage or loss of balance to the robot. These boundary conditions are incorporated into the constraints of the Lyapunov function to ensure that the results of subsequent calculations are physically feasible. Then, closed-loop feedback calculation and correction of the joint angles θ of the humanoid robot 10 (step A73) are performed. During the movement of the humanoid robot 10, real-time motion status messages from the IMU (Inertial Measurement Unit) 15 are continuously received, and the actual joint angles θ and angular velocities are compared with the expected values ​​calculated by the Lyapunov function to obtain the error value. In other words, based on the stability theory of the Lyapunov function, the joint angles θ are corrected through a closed-loop feedback control algorithm. Here, the core of the closed-loop feedback control in step A73 is to dynamically adjust the control input according to the magnitude and direction of the error, so that the error gradually decreases and approaches zero. Through continuous iterative calculation and adjustment, the joint angles θ gradually approach the stable target value obtained by the Lyapunov function. Finally, based on the calculated correction results, the femur assembly 11 and tibia assembly 12 are driven (step A74). This continuously optimized and corrected closed-loop control process enables the humanoid robot 10 to achieve stable balance in its gait. Thus, by utilizing the Lyapunov stability function, combined with precise dynamic information acquisition, algorithm calculation, and feedback control, the humanoid robot 10 can effectively achieve stable balance in its bipedal gait, quickly recovering and maintaining stable balance even under complex and unpredictable ground conditions.

[0056] Therefore, the humanoid robot 10 of this invention features a specially designed foot assembly 14, which can adapt to complex terrains such as uphill, downhill, undulating, and uneven surfaces. It also solves the problem of excessive energy consumption during walking caused by the lack of energy storage and impact absorption mechanisms in the rigid foot structure of bipedal robots. Furthermore, the posture control method of this humanoid robot 10, based on the analysis and calculation of the MPC model predictive control algorithm, can effectively perform data iteration, feedback control, and real-time adjustment of the bipedal robot's gait, thereby achieving the effects of predicting, planning, and optimizing the gait. It also enables autonomous balance and posture adjustment under various motion conditions such as starting from a standstill, deceleration during movement, uphill, and downhill. Even when facing uneven, complex, and unpredictable terrain, the humanoid robot 10 can autonomously return to stable balance through iteration and error correction.

[0057] The present invention has been described above with reference to embodiments, but it is not intended to limit the scope of the patent rights claimed by the present invention. The scope of patent protection shall be determined by the appended patent claims and their equivalent fields. Any modifications or refinements made by those skilled in the art without departing from the spirit or scope of this patent shall be considered equivalent changes or designs made under the spirit disclosed in this invention and shall be included within the scope of the patent application described below.

Claims

1. A humanoid robot comprising a torso assembly (13), two femur assemblies (11), two tibialis assemblies (12), two footplate assemblies (14), and a plurality of inertial measurement units (15), wherein the femur assemblies (11) are connected above the torso assembly (13) and pivotally connected below the tibialis assemblies (12), and the tibialis assemblies (12) are connected below the footplate assemblies (14), characterized in that: At least one inertial measurement unit (15) is provided on the torso assembly (13), the femur assembly (11), and the tibia assembly (12) to detect the movement and rotational inertia of various parts of the humanoid robot (10). The foot assembly (14) includes a first ground contact assembly (141), a second ground contact assembly (142), a third ground contact assembly (143), a fourth ground contact assembly (144), multiple pivot assemblies (145), and multiple pressure sensing units (146). The first, second, third, and fourth ground contact assemblies (144) are provided with pressure sensing units (146) facing the ground or adjacent to the periphery of the ground. The first, second, third, and fourth ground contact components (144) are connected in series via the pivot assembly (145). The first ground contact component (141) is located in the forward direction (FD) of the torso assembly (13), and the fourth ground contact component (144) is located in the rearward direction (BD) of the torso assembly (13). The fourth ground contact component (144) is connected to the lower side of the shinbone assembly (12) via a pivot assembly (145). The humanoid robot (10) performs ZMP zero-moment point gait control and analysis through the inertial measurement unit (15) and performs center of gravity balance feedback correction through the pressure sensing unit (146).

2. The humanoid robot according to claim 1, characterized in that: The femoral bone assembly (11) is provided with two inertial measurement units (15), which are respectively located at the two axial ends of the femoral bone assembly (11); or, the calf bone assembly (12) is provided with two inertial measurement units (15), which are respectively located at the two axial ends of the calf bone assembly (12).

3. The humanoid robot according to claim 1, characterized in that: At least one inertial measurement unit (15) is provided on the first ground contact assembly (141), the second ground contact assembly (142), the third ground contact assembly (143) or the fourth ground contact assembly (144) for detecting its movement and rotational inertia.

4. The humanoid robot according to claim 3, characterized in that: The inertial measurement unit (15) of the first ground contact assembly (141), the second ground contact assembly (142), the third ground contact assembly (143), or the fourth ground contact assembly (144) is disposed on the pivot assembly (145) or facing the ground.

5. The humanoid robot according to claim 1, characterized in that: At least one buffer rod (147) extends from the third ground contact assembly (143) and is connected to the second ground contact assembly (142).

6. The humanoid robot according to claim 1, characterized in that: The second grounding assembly (142) extends in the rearward direction (BD) and is connected to the fourth grounding assembly (144) via a pivot assembly (145).

Images