Balance control method for humanoid robot on complex terrain based on state and disturbance observer

Abstract

The application provides a complex terrain balance control method for humanoid robots based on a state and disturbance observer, and belongs to the technical field of robot application. The problem of instability of a humanoid robot when walking in a complex environment is solved. The technical scheme comprises the following steps: step 1, a simplified humanoid robot "spring-damping" inverted pendulum model is established; step 2, a state and disturbance observer is designed; step 3, based on step 2, a state feedback controller with the real-time observation object of the state and disturbance observer as input is designed; step 4, a suitable controller pole and controller observation point are selected; and step 5, simulation verification of the posture balance control of the humanoid robot in a complex terrain is performed. The application has the beneficial effect that, especially in disaster rescue and complex environment operation, the robot's robustness to interference and task execution capability are improved.

Description

Technical Field

[0001] The present invention relates to the field of robot application technology, and in particular to a complex terrain balance control method for a humanoid robot based on a state and disturbance observer. Background Art

[0002] In the field of modern robotics, humanoid robots, due to their flexibility and adaptability, have become a key research area in service robotics, medical rehabilitation, and complex industrial operations. However, due to their high degrees of freedom in joints and complex kinematic structures, humanoid robots often face unpredictable external environments and uncertainties in their own system parameters. Unexpected external disturbances (such as collisions and uneven surfaces) can lead to degraded control system performance and even robot instability. Therefore, improving the posture control performance of humanoid robots and ensuring the coordination and stability of the entire system are technical challenges in humanoid robotics research.

[0003] The paper "PID-Based Balance Control of Humanoid Robots in Dynamic Environments" designs a posture control method based on a PID controller to meet the needs of humanoid robots maintaining posture balance in dynamic environments. This method has two shortcomings: (1) PID control may not be able to effectively maintain posture balance when facing large external disturbances; (2) it has certain limitations when dealing with complex nonlinear systems and has difficulty coping with coupling effects and dynamic changes in multi-degree-of-freedom systems.

[0004] The paper "Fuzzy Logic Control of Bipedal Robot Posture Stability" designs a fuzzy logic-based posture control method to meet the need for bipedal robots to maintain posture stability in complex environments. This method has two shortcomings: (1) Fuzzy logic control relies on a complex set of rules. When the complexity of the system increases, the number of fuzzy logic rules will also increase significantly, resulting in increased computational complexity and affecting real-time performance. (2) Fuzzy rules are designed based on specific environmental conditions. When the external environment changes significantly, the adaptability of fuzzy logic control may be insufficient, affecting the stability of posture control.

[0005] How to solve the above technical problems is the subject faced by the present invention. Summary of the Invention

[0006] The purpose of the present invention is to provide a complex terrain balance control method for a humanoid robot based on a state and disturbance observer, aiming to improve the stability of the humanoid robot when walking in complex environments, especially its adaptability in disaster rescue and complex working environments. By designing a dual-action state and disturbance observer, the controller realizes real-time monitoring and adjustment of the robot's dynamic posture and external interference, ensuring that the robot can maintain balance and perform tasks in various uncertain environments.

[0007] The inventive concept of the present invention is as follows: taking into account that the impact force generated when the robot walks on unknown uneven ground will affect its posture balance, the present invention is inspired by the traditional inverted pendulum model and establishes a simplified "spring-damper" inverted pendulum model of the humanoid robot; then, a state and disturbance dual-action observer is designed to observe in real time the inclination angle, angular velocity and continuous disturbance of the center of mass of the humanoid robot during walking; secondly, based on the observer, a state feedback controller is designed with the real-time observation objects of the state and disturbance observer as input to maintain the posture balance of the robot when it is subjected to unknown disturbances in complex terrain; finally, according to the different stages of the humanoid robot's walking process, the present invention designs a variety of controller schemes, and through system identification and appropriate selection of poles, determines the optimal model parameters to ensure that the robot can maintain a stable balance posture in various environments.

[0008] In order to achieve the above-mentioned object of the invention, the present invention adopts a technical solution specifically as follows: a complex terrain balance control method of a humanoid robot based on a state and disturbance observer, comprising the following steps:

[0009] Step 1: Considering that the impact force generated when the robot walks on an unknown uneven surface will affect its posture balance, a simplified humanoid robot "spring-damper" inverted pendulum model is established, inspired by the traditional inverted pendulum model. This model can effectively absorb the negative effects of impact;

[0010] Step 2: Design a state and disturbance observer for the humanoid robot that is subject to unpredictable interference in complex terrain. This observer is used to observe the inclination angle, angular velocity, and continuous disturbance of the center of mass of the humanoid robot in real time during walking.

[0011] Step 3: Based on step 2, a state feedback controller is designed that uses the real-time observation objects of the state and disturbance observer as input to maintain the posture balance of the robot when walking on complex terrain and being subjected to unknown disturbances;

[0012] Step 4: For the observer-based state feedback controller designed above, determine the specific model parameters of the humanoid robot's walking process through system identification, select appropriate controller poles and controller observation points, and divide the model into the coronal plane model and sagittal plane model for the single support stage, and the coronal plane model and sagittal plane model for the double support stage according to the robot's walking phase and overall structure;

[0013] Step 5: Based on the Aelos humanoid standard platform, conduct simulation verification of the posture balance control of the humanoid robot in complex terrain to evaluate the actual performance of the posture balance controller based on the state and disturbance observer.

[0014] Furthermore, in step 1, the motion equation of the simplified humanoid robot "spring-damper" inverted pendulum model is as follows:

[0015]

[0016] Where m is the mass, l is the length of the connecting rod, c is the damping constant, k is the spring constant, g is the acceleration due to gravity, and θ is the swing angle. is the angular velocity of the pendulum, is the angular acceleration, u θ is the input angle of the inverted pendulum; if θ and u θ is very small, sinθ≈θ and lu θ ≈u, then the equation of motion can be approximated as follows:

[0017]

[0018] Furthermore, in step 2, it is assumed that the disturbance F d , the input is u, and the state space equation is:

[0019]

[0020] The above equation can be simply expressed as follows:

[0021]

[0022] y=C′z+D′u

[0023] in:

[0024]

[0025] Therefore, the observer is designed as follows:

[0026]

[0027] Where L is the observer gain determined by the pole placement method.

[0028] Furthermore, in step 3, if the state is known, a state feedback controller can be used to maintain balance, and the formula is as follows:

[0029]

[0030] Where K is the control gain determined by the pole placement method, r is the reference input value, N is a constant, α is the total gain of the controller change effect, and COM ref is the desired center of mass position. The transfer function between r and θ is as follows:

[0031]

[0032] Furthermore, in step 3, in order to maintain the posture balance of the humanoid robot, an observer-based state feedback controller is used, and the input u is defined as follows:

[0033]

[0034] Where, is the estimated state quantity, is the estimated disturbance. The state feedback controller can effectively eliminate the estimated disturbance, and the formula is as follows:

[0035]

[0036] Furthermore, in step 4, the humanoid robot has a single-support stage and a double-support stage during walking, and the model of the humanoid robot in each stage is different in the sagittal plane and the coronal plane, so the specific parameters of the four models need to be determined through system identification.

[0037] Furthermore, in step 4, the analytical solution of ZMP is:

[0038]

[0039] in:

[0040]

[0041] Assume the peak values ​​of ZMP output are P1 and P2,

[0042] P1=(t1,y1)

[0043] P2=(t2,y2)

[0044] The damping constant and spring constant are calculated as follows:

[0045]

[0046] The observer poles and controller poles are appropriately selected by trial and error.

[0047] Furthermore, in step 4, the force / torque sensor on the foot detects phase changes during the walking process of the humanoid robot. When the robot enters a new phase, such as from the double support phase to the single support phase or from the single support phase to the double support phase, a new controller is replaced, and the total gain α of the new controller is new Increase smoothly from 0 to 1t in a short time sh , and the total gain α of the previous controller pre By reducing from 1 to 0, the conversion can be completed instantly, which effectively improves the stability of the humanoid robot when changing its posture.

[0048] Compared with the prior art, the present invention has the following beneficial effects:

[0049] 1. By designing a posture balance controller based on a state and disturbance observer, the present invention can estimate and eliminate the influence of external disturbances in real time. Compared with traditional PID control and fuzzy logic control, the present invention significantly improves the stability of the robot's posture balance when facing external disturbances, especially in complex and dynamic environments, and has stronger robustness.

[0050] 2. In response to the multi-degree-of-freedom and nonlinear problems of humanoid robots during walking, the present invention designs multiple controller solutions for different environments and parameters, which can adaptively adjust the controller parameters and effectively maintain the stability of the robot's posture when the system parameters change or the environment changes suddenly, overcoming the limitations of traditional methods in dealing with multi-degree-of-freedom coupling effects.

[0051] 3. Compared with fuzzy logic control that relies on a large number of rules, the present invention realizes a simpler and more efficient control method by designing an attitude balance controller based on state and disturbance observer. While ensuring control accuracy, it greatly reduces the amount of calculation, improves the real-time performance and operating efficiency of the controller, and is suitable for highly dynamic working environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] The accompanying drawings are used to provide further understanding of the present invention and constitute a part of the specification. They are used to explain the present invention together with the embodiments of the present invention and do not constitute a limitation of the present invention.

[0053] Figure 1 This is a workflow diagram of the humanoid robot complex terrain balance controller based on the state and disturbance observer of the present invention.

[0054] Figure 2 This is a mathematical model diagram of the "spring-damping" inverted pendulum in the present invention.

[0055] Figure 3 This is a system structure diagram of the humanoid robot in the present invention.

[0056] Figure 4 This is a gain scheduling diagram for the controller to smoothly transition when the walking phase of the humanoid robot moves from the double-support stage to the single-support stage in the present invention.

[0057] Figure 5 This is a physical picture of the AELOS SMART robot under the AELOS humanoid standard platform of the present invention.

[0058] Figure 6 The natural response curve diagram of the present invention shows the angle change of the "spring-damped" inverted pendulum over time without control input.

[0059] Figure 7 The observer performance curve in the present application; the angle of the observer estimate is compared with the true angle.

[0060] Figure 8 The closed-loop response curve in the present application; the system response under the state feedback controller is shown. DETAILED DESCRIPTION

[0061] In order to make the purpose, technical solutions and advantages of the present application more clear and explicit, the present application is further described in detail below in combination with the drawings and examples. Of course, the specific embodiments described herein are only used to explain the present application and not to limit the present application.

[0062] Example 1

[0063] Referring to Figures 1 to 8 By comparing with the conventional controller without the designed observer, the control effect of the present embodiment on the posture balance of the humanoid robot under complex terrain is comprehensively shown.

[0064] In order to test the performance of different control methods, four kinds of controllers are designed respectively: controller A (conventional PID controller), controller B (fuzzy logic controller), controller C (LQR controller), and controller D (humanoid robot complex terrain balance controller based on state and disturbance observer). During the experiment, various disturbance conditions are set, including random impact force and uneven terrain, the purpose is to test the response ability of the controller in the face of uncertain environment. The following performance indicators are mainly investigated: posture error (angle and position error deviating from the target posture), steady-state error (residual error after system stabilization) and control signal energy (the strength of the control input required by the controller to maintain balance), and the following controller performance comparison table 1 is obtained:

[0065] Table 1 Comparison of controller performance

[0066]

[0067] The experimental results show that the present embodiment performs excellently in all indicators, especially in the face of sudden disturbance, the controller D can recover balance in the shortest time, the posture error and steady-state error are the smallest, and the control signal energy is also relatively low. In addition, the controller C is the second, which also shows good response speed and relatively low error, while the controller A and the controller B recover slowly and have relatively large error in the case of strong interference.

[0068] From the perspective of controller implementation, this embodiment can capture system state and disturbance information in real time and effectively eliminate external interference using a state feedback control algorithm, significantly improving the system's robustness and control accuracy. In contrast, traditional PID and fuzzy controllers, while simple, are prone to large control errors when faced with complex environments and strong disturbances. The LQR controller outperforms PID and fuzzy controllers in terms of control signal energy and state feedback control, but still suffers from large errors.

[0069] Overall, the aforementioned traditional control methods all have certain limitations when adjusting the posture and balance of humanoid robots. In particular, when responding to external interference, the control effectiveness and system robustness of these traditional control methods still need to be improved compared to those of this embodiment. These comparative results provide an important basis for verifying the effectiveness of this embodiment's method in subsequent embodiments.

[0070] Example 2

[0071] See also Figures 1 to 8 , the technical solution provided by this embodiment is as follows: Figure 5 As shown, using the AELOS humanoid standard platform and the AELOS SMART robot model, this embodiment proposes a complex terrain balance controller for a humanoid robot based on a state and disturbance observer to achieve posture balance control of the humanoid robot in complex environments. First, a simplified "spring-damper" inverted pendulum model of the humanoid robot is established. Second, a state and disturbance dual-action observer is designed to observe the inclination angle, angular velocity, and continuous disturbances of the humanoid robot's center of mass in real time during walking. Then, based on the observer, a state feedback controller is designed, setting the robot's zero torque point as the output and the inclination angle, angular velocity, and disturbances of the center of mass as the state. By utilizing the posture and disturbance information provided by the observer, the controller can dynamically adjust the robot's control input to ensure the robot's balance in complex environments. Finally, multiple controller schemes are designed based on the humanoid robot's motion characteristics in different environments (such as the parameters or movement direction of different legs). Through system identification, the optimal controller parameters are determined to ensure that the robot can maintain balance in all walking phases.

[0072] The method flow of this embodiment is as follows Figure 1 The specific method is as follows:

[0073] 1. Considering the impact force generated when the robot walks on an unknown uneven surface, a simplified "spring-damper" inverted pendulum model of a humanoid robot is established based on the traditional inverted pendulum model to effectively absorb the negative effects of the impact.

[0074] The robot used in this example uses high-gain position control for all actuators, taking into account the damping effects from the mechanical hardware, joint controllers, and the sole material. The equation of motion for the simplified humanoid robot "spring-damper" inverted pendulum model is as follows:

[0075]

[0076] Where m is the mass, l is the length of the connecting rod, c is the damping constant, k is the spring constant, g is the acceleration due to gravity, and θ is the swing angle. is the angular velocity of the pendulum, is the angular acceleration, u θ is the input angle of the inverted pendulum. If θ and u θ is very small, sinθ≈θ and lu θ ≈u, then the equation of motion can be approximated as follows:

[0077]

[0078] like Figure 2 As shown, the forces and moments at the ankle joint are measured and the ZMP of the robot along the x-axis is calculated as follows:

[0079]

[0080] Among them F z and M y is the z-direction ground reaction force and y-direction ground reaction force measured by the FT sensor at the ankle. y is also the reaction torque of the spring and damper, so it can be written in terms of the action of the spring and damper as:

[0081]

[0082] Furthermore, if the z-direction motion is small, it can be assumed that F z The equation for ZMP is the same as the weight of the robot.

[0083]

[0084] 2. For humanoid robots that are subject to unpredictable interference in complex terrain, a state and disturbance observer is designed to observe the inclination angle, angular velocity and continuous disturbance of the center of mass of the humanoid robot in real time during walking.

[0085] Overall exercise program such as Figure 3 When the desired ZMP is planned, the pattern generator generates the motion reference trajectory of COM and swing foot based on the online preview control. Assuming the disturbance F d , the state space equation is:

[0086]

[0087] The above equation can be simply expressed as:

[0088]

[0089] y=C′z+D′u

[0090] in:

[0091]

[0092] Therefore, the observer is designed as follows:

[0093]

[0094] Where L is the observer gain determined by the pole placement method.

[0095] 3. Based on step 2, design a state feedback controller. The proposed controller is placed Figure 3 In the dotted box, the control input of the robot is adjusted according to the estimated posture and disturbance to maintain the balance of the robot. The feedback controller formula for maintaining the balance state is as follows:

[0096]

[0097] Where K is the control gain determined by the pole placement method, r is the reference input value, N is a constant, α is the total gain of the controller change effect, and COM ref is the desired center of mass position. The transfer function between r and θ is as follows:

[0098]

[0099] A state feedback controller based on a state observer is used, and the input u is defined as follows:

[0100]

[0101] The state feedback controller can effectively eliminate the estimated disturbance, and the formula is as follows:

[0102]

[0103] 4. Based on step 3, for the humanoid robot with different modeling parameters for several legs and / or movement directions, perform system identification for four situations and design corresponding controllers.

[0104] Humanoid robots have single-support and double-support phases during walking. Furthermore, the humanoid robot model differs in the sagittal and coronal planes, so the observer and controller were designed for each plane. System identification was performed on the four models to determine the damping and spring constants for the simple model. The controller and observer poles were appropriately selected through trial and error, as shown in Table 2.

[0105] Table 2 Different control parameters of four models

[0106]

[0107] During walking, when the robot enters a new phase, such as from double support phase to single support phase or from single support phase to double support phase, the controller needs to be replaced. This phase change is detected by the force / torque sensor on the foot. To achieve instantaneous phase transition, the total gain α of the new controller is new Increase smoothly from 0 to 1t in a short time sh , and the total gain α of the previous controller pre Decrease from 1 to 0, such as Figure 4 shown.

[0108] 5. Based on the Aelos humanoid standard platform, conduct simulation verification of the posture balance control of a humanoid robot in complex terrain, and evaluate the actual performance of the posture balance controller based on the state and disturbance observer.

[0109] First, a simplified humanoid robot "spring-damper" inverted pendulum model is established using MATLAB / Simulink. The state matrix A, input matrix B, output matrix C, and transfer matrix D of the system are defined through the state space model to capture the dynamic characteristics of the robot. The free response simulation is then used to verify the natural recovery ability of the system when there is no control input. Figure 6 As shown. Next, a dual-action state and disturbance observer is designed to estimate the inclination angle, angular velocity and continuous disturbance of the robot's center of mass in real time. The observer selects the appropriate gain L through the pole placement method, constructs the state space observer model, and simulates and verifies the observer's estimation accuracy of the system state, as shown in Figure 7 As shown. The results show that the observer can accurately track the true state and provide reliable state estimation for the feedback controller. Finally, based on the state information provided by the observer, the state feedback controller is designed, the control gain K is determined using the pole placement method, and a closed-loop control system is constructed, as shown in Figure 8 Simulation results show that the controller effectively improves the system's response speed and stability, enabling the humanoid robot to quickly regain balance when disturbed. Simulations of the closed-loop system's response validate the robustness of the state feedback controller, significantly improving posture control performance, particularly when dealing with dynamic changes in complex environments.

[0110] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A complex terrain balance control method for a humanoid robot based on a state and disturbance observer, characterized in that: The following steps are involved: Step 1: Considering that the impact force generated when the robot walks on an unknown uneven surface will affect its posture balance, a simplified "spring-damper" inverted pendulum model of a humanoid robot is established, inspired by the traditional inverted pendulum model. Step 2: Design a state and disturbance observer for the humanoid robot that is subject to unpredictable interference in complex terrain. This observer is used to observe the tilt angle, angular velocity, and continuous disturbance of the center of mass COM of the humanoid robot in real time during walking. Step 3: Based on step 2, a state feedback controller is designed that uses the real-time observation objects of the state and disturbance observer as input to maintain the posture balance of the robot when it is subjected to unknown disturbances in complex terrain; Step 4: For the observer-based state feedback controller designed above, determine the specific model parameters of the humanoid robot's walking process through system identification, select appropriate controller poles and controller observation points, and divide the model into the coronal plane model and sagittal plane model for the single support stage, and the coronal plane model and sagittal plane model for the double support stage according to the robot's walking phase and overall structure; Step 5: Based on the Aelos humanoid standard platform, conduct simulation verification of the posture balance control of the humanoid robot in complex terrain to evaluate the actual performance of the posture balance controller based on the state and disturbance observer.

2. The complex terrain balance control method of a humanoid robot based on a state and disturbance observer according to claim 1 is characterized in that: In step 1, the motion equation of the simplified humanoid robot "spring-damper" inverted pendulum model is as follows: Where m is the mass, l is the length of the connecting rod, c is the damping constant, k is the spring constant, g is the acceleration due to gravity, and θ is the swing angle. is the angular velocity of the pendulum, is the angular acceleration, u θ is the input angle of the inverted pendulum.

3. The complex terrain balance control method of a humanoid robot based on a state and disturbance observer according to claim 1 is characterized in that: In step 2, it is assumed that the disturbance F d , the input is u, and the disturbance is defined as an additional element of the state, then the state space equation is: The above equation can be expressed as follows: y=C′z+D′u in: Therefore, the observer is designed as follows: Where L is the observer gain determined by the pole placement method.

4. The complex terrain balance control method of a humanoid robot based on a state and disturbance observer according to claim 2 is characterized in that: In step 3, if the state is known, a state feedback controller is used to maintain balance. The formula is as follows: Where K is the control gain determined by the pole placement method, r is the reference input value, N is a constant, α is the total gain of the controller change effect, and COM ref For the desired center of mass position, the transfer function between r and θ is as follows:

5. The complex terrain balance control method of a humanoid robot based on a state and disturbance observer according to claim 4 is characterized in that: In step 3, in order to maintain the posture balance of the humanoid robot, an observer-based state feedback controller is used, and the input u is defined as follows: Where, is the estimated state quantity, is the estimated disturbance, and the state feedback controller eliminates the estimated disturbance. The formula is as follows:

6. The complex terrain balance control method of a humanoid robot based on a state and disturbance observer according to claim 1 is characterized in that: In step 4, the humanoid robot has a single-support stage and a double-support stage during walking, and the model of the humanoid robot in each stage is different in the sagittal plane and the coronal plane. Therefore, the specific parameters of the four models are determined through system identification, including the coronal plane model and the sagittal plane model of the single-support stage, and the coronal plane model and the sagittal plane model of the double-support stage.

7. The complex terrain balance control method of a humanoid robot based on a state and disturbance observer according to claim 5 is characterized in that: In step 4, the analytical solution of the zero moment point ZMP is: in: Assume the peak values ​​of ZMP output are P1 and P2, P1=(t1,y1) P2=(t2,y2) The damping constant and spring constant are calculated as follows: The observer poles and controller poles are appropriately selected by trial and error.

8. The complex terrain balance control method of a humanoid robot based on a state and disturbance observer according to claim 6 is characterized in that: In step 4, when the humanoid robot is walking, the force or torque sensor on the foot detects the phase change. When entering a new stage, from the double support stage to the single support stage or from the single support stage to the double support stage, the controller needs to be replaced in time. To ensure that the conversion is carried out instantly, the total gain α of the new controller is designed. new Increase smoothly from 0 to 1t in a short time sh , the total gain α of the previous controller pre Decrease from 1 to 0.

Images