Robot integrated joint disturbance compensation method based on double torque sensors and improved disturbance observer

Abstract

The invention relates to the technical field of robots, in particular to a robot integrated joint disturbance compensation method based on double torque sensors and an improved disturbance observer, and the method comprises the following steps: S1, building a joint dynamic model including a motor side, a harmonic reducer, a load side, an environment and manpower; s2, installing a torque sensor alpha between the bearing and the load, and measuring load inertia and external torque; a torque sensor beta is installed on a rigid wheel of the harmonic reducer, the torque of the rigid wheel is measured, the torque relation between the two sensors is deduced, and disturbance converted from the load side to the motor side is further obtained; s3, a double-torque sensor is combined with a motor side disturbance observer, the double-torque sensor and a feedback controller form a closed-loop control system of the integrated joint, the motor side disturbance observer observes and obtains motor side disturbance, disturbance converted from the load side to the motor side is obtained through derivation in the step S2, and then the two kinds of disturbance are compensated in a control loop; and S4, performing robust stability analysis on the closed-loop control system. Through collaborative design of the double torque sensors and the improved disturbance observer, accurate disturbance suppression is achieved, and the problems that due to robot integrated joint multi-source disturbance, torque control precision is insufficient, operation stability is not high, and modeling is complex are solved.

Description

Technical Field

[0001] This invention relates to the field of robotics, and in particular to an integrated joint disturbance compensation method for robots based on dual torque sensors and an improved disturbance observer. Background Technology

[0002] With the widespread application of robotics in industrial manufacturing, medical rehabilitation, and human-robot collaboration, integrated joints, as the core drive and control unit of robots, have become key factors determining robot performance in terms of control precision, dynamic response speed, and human-robot interaction safety. Especially for devices such as lightweight robotic arms and lower limb exoskeletons that require frequent interaction with the human body or environment, stringent requirements are placed on the torque tracking accuracy and reverse drive performance of the joints. These joints must not only achieve rapid torque response during forward drive but also possess low resistance characteristics during reverse drive (zero torque control), allowing for easy load manipulation and ensuring the flexibility and safety of human-robot collaboration.

[0003] However, the transmission system of a robot's integrated joint, such as motors and harmonic reducers, is susceptible to multi-source disturbances. These disturbances are widely distributed on the motor side, reduction gears, and load side, mainly manifesting as motor friction, gear meshing losses, bearing friction, and nonlinear hysteresis. In torque control systems, such disturbances lead to output torque loss, which not only hinders the joint's rapid tracking of the desired torque and reduces dynamic response performance, but also increases the operating resistance during reverse drive, severely impacting the human-machine interaction experience.

[0004] To address the aforementioned disturbance issues, existing technologies primarily employ three approaches: First, friction model-based compensation methods, which compensate for disturbances by establishing a friction dynamics model. However, joint friction is sensitive to temperature and load changes, and model parameters are prone to drift, leading to unstable compensation accuracy. Second, the single disturbance observer (DOB) method, which compensates by estimating lumped disturbances. However, traditional DOBs are highly dependent on the system model, require time-consuming parameter identification, and suffer from phase lag in the filters, making it difficult to balance forward response speed and reverse drive performance. Third, single torque sensor feedback schemes, which acquire torque information solely through load-side or motor-side sensors, cannot comprehensively capture multi-source disturbances. In particular, they easily overlook frictional losses at high-torque ports on the load side, resulting in incomplete disturbance compensation.

[0005] Furthermore, under complex working conditions such as heavy loads and high stiffness environments, existing solutions lack robustness, leading to increased torque tracking errors and prolonged settling times, making it difficult to meet the demands of high-precision collaboration. Therefore, how to overcome the reliance on models in traditional methods, achieve comprehensive perception and accurate compensation of multi-source disturbances, and simultaneously improve the dynamic response speed, steady-state accuracy, and reverse drive performance of joints under complex conditions has become a pressing technical challenge in the field of integrated robot joint control. To address this, we propose a robot integrated joint disturbance compensation method based on dual torque sensors and an improved disturbance observer. Summary of the Invention

[0006] Based on the technical problems existing in the background technology, this invention proposes a robot integrated joint disturbance compensation method based on dual torque sensors and an improved disturbance observer. Through the collaborative design of dual torque sensors and an improved disturbance observer, the method achieves precise disturbance suppression and solves the problems of insufficient torque control accuracy, low operational stability and complex modeling caused by multi-source disturbances in robot integrated joints.

[0007] This invention provides the following technical solution: a robot integrated joint disturbance compensation method based on dual torque sensors and an improved disturbance observer, comprising the following steps:

[0008] S1. Establish a joint dynamics model that includes the motor side, harmonic reducer, load side, environment, and human factors;

[0009] S2. Install torque sensor α between the bearing and the load to measure the load inertia and external torque; install torque sensor β on the rigid wheel of the harmonic reducer to measure the rigid wheel torque, derive the torque relationship between the two sensors, and further obtain the disturbance from the load side to the motor side.

[0010] S3. Combine the dual torque sensor with the motor-side disturbance observer and form an integrated closed-loop control system with the feedback controller. The motor-side disturbance observer observes the motor-side disturbance, and the disturbance from the load side to the motor side is derived through step S2. Then, the two disturbances are compensated in the control loop.

[0011] Preferably, in the joint dynamics model of step S1, and These represent the moment of inertia and damping coefficient of the motor, respectively. and These represent the moment of inertia and damping coefficient of the load, respectively. and These represent the damping coefficient and stiffness coefficient of the flexible wheel, respectively. Indicates the transmission ratio;

[0012] This assumes that the frictional torques on the motor side, load side, and harmonic drive are all linear; the joint is considered a dual-input system, and the motor torque ( As the command input, the external torque given by the person ( ) or contact torque ( As an external disturbance input; due to energy transfer losses between the wave generator and the flexspline, the transmission efficiency of harmonic drive is used as... express, For environmental stiffness coefficient, and These represent the angular positions on the motor side and the load side, respectively. Indicates angular velocity. If angular acceleration is represented, then the joint dynamics model can be expressed as:

[0013] ;

[0014] ;

[0015] ;

[0016] .

[0017] Preferably, the dynamic model includes motor dynamics, harmonic reducer dynamics, load dynamics, and environmental stiffness, which are expressed as follows: , , and , For the Laplace operator.

[0018] Preferably, the torque sensor α in step S2 measures the load inertia and external torque, as shown in the following formula:

[0019] ;

[0020] For sensor β, which measures the torque of the rigid wheel, the ideal torque balance between the components is described as follows:

[0021] ;

[0022] in, Let represent the torque of the rigid wheel, without considering the direction of the torque. This includes the torque loss caused by gear meshing. Therefore, the meshing efficiency between the rigid wheel and the flexible wheel can be expressed as: The relationship between the two torque sensors is derived as follows:

[0023] ;

[0024] The torque due to friction on the load side and gear meshing loss can be calculated as follows:

[0025] ;

[0026] Here, the measured values ​​of torque sensors α and β are denoted as follows: and The calculation results represent the load-side disturbance; to achieve friction compensation, the disturbance value from the load side to the motor side is as follows:

[0027] .

[0028] Preferably, in step S3, the measured values ​​from the two torque sensors are combined with those from the motor-side disturbance observer to compensate for overall friction or disturbance, thus fusing the torque measurements. Motor side angle position and motor current To estimate the torque disturbance on the motor side; simultaneously, at the low-speed port, the energy loss between the rigid wheel and the flexible wheel ( The sum of friction and load is uniformly represented as Using this disturbance observer, the torque disturbance is estimated as follows:

[0029] .

[0030] Preferably, the method further includes step S4: performing robust stability analysis on the closed-loop control system.

[0031] Preferably, in step S4, the model uncertainty is treated as multiplicative fluctuation. Defined as:

[0032] ;

[0033] in, and They represent from arrive The actual and nominal models of the open-loop transfer function are represented by the product of the transfer functions of the system's forward path:

[0034] ;

[0035] in, For a feedback controller, choose a PI type: , Indicates when the load comes into contact with the environment. arrive The nominal model open-loop transfer function; in addition, Defined as the closed-loop transfer function with a nominal model:

[0036] ;

[0037] The above equation represents the complementary sensitivity function, which will be used here. Redefining According to the small gain theorem, the robust stability condition of the system is expressed by the following equation:

[0038] ;

[0039] Closed-loop control system must meet To achieve robust stability.

[0040] This invention provides a robot integrated joint disturbance compensation method based on dual torque sensors and an improved disturbance observer. A torque sensor is installed on both the rigid wheel and the load end of the harmonic reducer. Using measured data from these two sensors, disturbance relationships are established through theoretical derivation, directly acquiring load-side disturbance information (including load-side bearing friction, gear meshing losses, etc.) without relying on complex dynamic models involving nonlinear friction. Simultaneously, by integrating the signals from the sensors mounted on the rigid wheel with a simplified motor model, an improved disturbance observer is designed to accurately estimate motor-side disturbances (such as motor friction and losses before harmonic drive deceleration). The disturbance estimation results are input into the disturbance compensator, forming an integrated closed-loop control system for the joint with a feedback controller, ultimately achieving high-precision torque tracking control. This design eliminates complex nonlinear friction modeling steps, avoids errors caused by model mismatch and parameter drift, and comprehensively observes and compensates for various disturbances originating from the motor side, harmonic reducer, and load side, significantly optimizing the joint's torque control dynamic response and steady-state accuracy. Attached Figure Description

[0041] Figure 1 This is a schematic diagram of the integrated joint dual inertia model of the present invention;

[0042] Figure 2 This is a structural diagram of the closed-loop control system model of the present invention;

[0043] Figure 3 Diagram of a traditional disturbance observer and friction compensation structure;

[0044] Figure 4 This is a diagram showing the installation position of the torque sensor of the present invention;

[0045] Figure 5 This is a model diagram of the disturbance observer combined with a torque sensor according to the present invention.

[0046] In the diagram: 1. Rigid wheel; 2. Flexible wheel; 3. Wave generator; 4. Torque motor; 5. Motor-side encoder; 6. Torque sensor α; 7. Torque sensor β; 8. Load; 9. Load-side encoder. Detailed Implementation

[0047] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0048] This invention provides a technical solution: a robot integrated joint disturbance compensation method based on dual torque sensors and an improved disturbance observer, comprising the following steps:

[0049] S1. Establish a basic joint dynamics model (including the motor side, harmonic reducer, load side, environment, and human factors) to provide a theoretical basis for subsequent analysis. Here, friction, which is usually nonlinear, is treated as linear, so there is no need to establish a complex nonlinear model for friction. Subsequent steps will compensate for nonlinear friction.

[0050] S2. Install torque sensor α between the bearing and the load to measure load inertia and external torque; install torque sensor β on the rigid wheel of the harmonic reducer to measure the rigid wheel torque. Derive the torque relationship between the two sensors, and further derive the disturbance from the load side to the motor side.

[0051] S3. Combine the dual torque sensor with the motor-side disturbance observer. The motor-side disturbance observer observes the motor-side disturbance (including various nonlinear frictions). The disturbances referred from the load side to the motor side (including nonlinear frictions and gear meshing losses) are derived in step 2. Then, these two disturbances are compensated in the control loop, such as... Figure 5 As shown.

[0052] S4. Describe the method for judging robust stability. When a system comes into contact with its environment, changes in the stiffness of the environment are considered as uncertainties in the system; therefore, robust stability analysis must be performed on the system.

[0053] The integrated joint dynamics model is established in step S1. The transmission components of the integrated joint are typically composed of harmonic reducers. Since the stiffness of the harmonic reducer is relatively low compared to other components in the joint, it is often approximated as a spring-damped system. Therefore, the integrated joint can be considered as a series elastic actuator (SEAs). The integrated joint can be simplified to a two-inertia model, such as... Figure 1 As shown, the harmonic reducer structure mainly consists of a wave generator, a flexspline, and a circular spline.

[0054] In this model, and These represent the moment of inertia and damping coefficient of the motor, respectively. and These represent the moment of inertia and damping coefficient of the load, respectively. and These represent the damping coefficient and stiffness coefficient of the flexible wheel, respectively. This represents the transmission ratio. Friction is usually modeled nonlinearly, but for ease of system analysis, it is assumed here that the frictional torques on the motor side, load side, and harmonic drive are all linear. The joint is considered a dual-input system, with the motor torque ( As the command input, the external torque given by the person ( ) or contact torque ( (This is used as an external disturbance input.) Due to energy transfer losses between the wave generator and the flexspline, the transmission efficiency of harmonic drive is... express, For environmental stiffness coefficient, and These represent the angular positions on the motor side and the load side, respectively. Indicates angular velocity. Let angular acceleration be the metric, then the dynamics of the system can be expressed as:

[0055] (1)

[0056] (2)

[0057] (3)

[0058] (4)

[0059] in, This indicates the output torque of the harmonic reducer.

[0060] The system consists of four parts: motor dynamics, harmonic reducer dynamics, load dynamics, and environmental stiffness, which are expressed as follows: , , and , For the Laplace operator, such as Figure 2 As shown. Among them. and These represent the motor current and torque constant, respectively.

[0061] Harmonic reducers experience power loss during transmission due to friction. This energy loss primarily originates from friction in four key components: the wave generator bearing, the flexible bearing, the gear meshing, and the output bearing. The friction generated by the wave generator bearing before reduction can be categorized as damping on the motor side. Energy loss exists between the wave generator and the flexible gear (denoted as...). This loss is mainly caused by the friction between the outer ring of the flexible bearing of the wave generator and the inner wall of the flexible wheel. Since this loss still exists at the high-speed transmission port, it can be considered together with the motor friction. This combined friction will be referred to as [missing information - likely a typo]. Load-side friction originates from the relative motion between the load and the bearing. In some scenarios, if a support bearing is installed between the flexible wheel and the joint base, the friction in this area is also classified as load-side friction. These types of load-side friction are collectively denoted as... .

[0062] Traditional disturbance observers such as Figure 3 As shown. Traditional disturbance observers are generally used to observe disturbances on the motor side (including nonlinear friction), and cannot provide a comprehensive observation of disturbances in the entire system. Observing disturbances on both the motor side and the load side at the same time would undoubtedly increase the workload considerably.

[0063] Traditional disturbance observation methods typically focus more on motor-side friction, including friction between the wave generator and the motor, as well as unknown disturbances, while output bearing friction (i.e., load-side disturbances) is often neglected. The disturbance observer includes the same Q-filter (designed as a second-order low-pass filter). This indicates the cutoff frequency of the filter, and its design follows standard empirical tuning methods. and Indicates the nominal model parameters, and estimates the disturbance torque. Represented as:

[0064] (5)

[0065] Estimated disturbance torque This includes the already aggregated motor-side friction and other disturbance torques originating from harmonic drives (denoted as...). This compensation effectively mitigates disturbances and improves forward drive performance. However, it should be noted that the disturbance estimate also includes external torque, which will be compensated for during the control process. Therefore, using the traditional disturbance observer control method in a dual-inertia system will reduce reverse drive performance.

[0066] In step S2, the torque sensor is installed and the load-side disturbance is analyzed. Due to the characteristics of the joint, perfect control performance on the motor side cannot guarantee perfect control performance on the load side. In practical applications, load-side disturbances also have a significant impact on the performance of the integrated joint, so disturbance compensation must also be considered on the load side. In the integrated joint system, the frictional torque on the load side is mainly generated from the output bearing, so a torque sensor α (labeled 6 in the figure) is installed between the bearing of the flexspline 2 and the load 8. To obtain the disturbance, another torque sensor β (labeled 7 in the figure) is installed on the fixed side. Due to the rotational characteristics of the flexspline 2, it is difficult to directly install the torque sensor on the flexspline 2, so it is installed on the rigid wheel 1 of the harmonic reducer (labeled β). This also includes the torque motor 4, the motor-side encoder 5, and the load-side encoder 9, as shown below. Figure 4 As shown.

[0067] Therefore, the torque sensor α measures the load inertia and external torque, as shown in equation (6):

[0068] (6)

[0069] It is the combined torque of the flexible wheel torque and the frictional torque. For sensor β, which measures the torque of the rigid wheel, the ideal torque balance between the components can be described as:

[0070] (7)

[0071] in, Let represent the torque of the rigid wheel, without considering the direction of the torque. This includes the torque loss caused by gear meshing. Therefore, the meshing efficiency between the rigid wheel and the flexible wheel can be expressed as: The relationship between the two torque sensors is derived as follows:

[0072] (8)

[0073] Therefore, the torque due to friction on the load side and gear meshing loss can be calculated as follows:

[0074] (9)

[0075] Here, the measured values ​​of torque sensors α and β are denoted as follows: and The calculation results represent the load-side disturbance. To achieve friction compensation, the disturbance values ​​from the load side to the motor side are as follows:

[0076] (10).

[0077] In step S3, the dual torque sensors are combined with the motor-side disturbance observer.

[0078] Based on the above analysis of load-side friction and gear meshing loss torque, they can be aggregated into a load-side disturbance, calculated by two torque sensors and converted to the motor side. However, motor-side friction is still compensated by a disturbance observer. This section combines the measurements from the two torque sensors with the motor-side disturbance observer to compensate for the overall friction or disturbance. The integrated control block diagram is shown below. Figure 5 As shown. Its design concept is to integrate torque measurement values. Motor side angle position and motor current To estimate the torque disturbance on the motor side. Meanwhile, at the low-speed port, the energy loss between the rigid and flexible gears ( It can also be combined with the load friction lumped and uniformly represented as Using this disturbance observer, the torque disturbance is estimated as follows:

[0079] (11);

[0080] In the above model, the disturbance observer estimates the motor side... This refers to the lumped disturbance at the high-speed port, including motor friction, friction between the wave generator and the flexible bearing, and friction between the flexible bearing and the inner ring of the flexspline; the measured value of the load-side disturbance is calculated using dual torque sensors. It incorporates the meshing loss torque between the rigid wheel and the flexible wheel, as well as the bearing friction on the load side of the low-speed port.

[0081] In step S4, the robust stability judgment method states that when the nominal model matches the actual model, i.e. At that time, Q filter The system must be stable (i.e., nominally stable). However, fluctuations in the controlled object due to parameter uncertainties can reduce the stability of the feedback system. Therefore, when designing the Q-filter, it is necessary to consider whether the feedback system can maintain stability (i.e., robust stability) when the object does not match the nominal model. In practical applications, the proposed method requires estimation of parameters from both motor sides. and These estimation errors are small and negligible. Here, system uncertainty is considered as the environmental stiffness when the load is in contact with the environment. The changes. Model uncertainty can be viewed as multiplicative fluctuations. Defined as:

[0082] (12)

[0083] in, and They represent from arrive The actual and nominal models of the open-loop transfer function can be represented by the product of the transfer functions of the system's forward paths (including disturbance observers):

[0084] (13)

[0085] in, For a feedback controller, choose a PI type: , Indicates when the load comes into contact with the environment. arrive The nominal model open-loop transfer function. Furthermore, Defined as the closed-loop transfer function with a nominal model:

[0086] (14)

[0087] Equation (14) represents the complementary sensitivity function, which will be used here. Redefining According to the small gain theorem, the robust stability condition of the system is expressed by the following equation:

[0088] (15)

[0089] Therefore, the control system must meet the following requirements. To achieve robust stability. In this invention, in order to evaluate the... Robustness to change can be compared in the control system. and .

[0090] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A robot joint perturbation compensation method based on dual torque sensors and an improved perturbation observer, characterized in that: Includes the following steps: S1. Establish a joint dynamics model that includes the motor side, harmonic reducer, load side, environment, and human factors; S2. Install torque sensor α between the bearing and the load to measure the load inertia and external torque; install torque sensor β on the rigid wheel of the harmonic reducer to measure the rigid wheel torque, derive the torque relationship between the two sensors, and further obtain the disturbance from the load side to the motor side. S3. Combine the dual torque sensor with the motor-side disturbance observer and form an integrated closed-loop control system with the feedback controller. The motor-side disturbance observer observes the motor-side disturbance, and the disturbance from the load side to the motor side is derived through step S2. Then, the two disturbances are compensated in the control loop.

2. The robot integrated joint disturbance compensation method based on dual torque sensors and an improved disturbance observer according to claim 1, characterized in that: In the joint dynamics model of step S1 and These represent the moment of inertia and damping coefficient of the motor, respectively. and These represent the moment of inertia and damping coefficient of the load, respectively. and These represent the damping coefficient and stiffness coefficient of the flexible wheel, respectively. Indicates the transmission ratio; This assumes that the frictional torques on the motor side, load side, and harmonic drive are all linear; the joint is considered a dual-input system, and the motor torque ( As the command input, the external torque given by the person ( ) or contact torque ( As an external disturbance input; due to energy transfer losses between the wave generator and the flexspline, the transmission efficiency of harmonic drive is used as... express, For environmental stiffness coefficient, and These represent the angular positions on the motor side and the load side, respectively. Indicates angular velocity, If angular acceleration is used, then the joint dynamics model can be expressed as: ； ； ； 。 3. The robot integrated joint disturbance compensation method based on dual torque sensors and an improved disturbance observer according to claim 2, characterized in that: The dynamic model includes motor dynamics, harmonic reducer dynamics, load dynamics, and environmental stiffness, which are expressed as follows: , , and , For the Laplace operator.

4. The robot integrated joint disturbance compensation method based on dual torque sensors and an improved disturbance observer according to claim 3, characterized in that: In step S2, the torque sensor α measures the load inertia and external torque, as shown in the following formula: ； For sensor β, which measures the torque of the rigid wheel, the ideal torque balance between the components is described as follows: ； in, Let represent the torque of the rigid wheel, without considering the direction of the torque. This includes the torque loss caused by gear meshing. Therefore, the meshing efficiency between the rigid wheel and the flexible wheel can be expressed as: The relationship between the two torque sensors is derived as follows: ； The torque due to friction on the load side and gear meshing loss can be calculated as follows: ； Here, the measured values ​​of torque sensors α and β are denoted as follows: and The calculation results represent the load-side disturbance; to achieve friction compensation, the disturbance value from the load side to the motor side is as follows: 。 5. The robot integrated joint disturbance compensation method based on dual torque sensors and an improved disturbance observer according to claim 4, characterized in that: In step S3, the measurements from the two torque sensors are combined with those from the motor-side disturbance observer to compensate for overall friction or disturbance, and the torque measurements are fused. Motor side angle position and motor current To estimate the torque disturbance on the motor side; simultaneously, at the low-speed port, the energy loss between the rigid wheel and the flexible wheel ( The sum of friction and load is uniformly represented as Using this disturbance observer, the torque disturbance is estimated as follows: 。 6. The robot integrated joint disturbance compensation method based on dual torque sensors and an improved disturbance observer according to claim 5, characterized in that: It also includes step S4: performing robust stability analysis on the closed-loop control system.

7. The robot integrated joint disturbance compensation method based on dual torque sensors and an improved disturbance observer according to claim 6, characterized in that: In step S4, the model uncertainty is treated as multiplicative fluctuation. Defined as: ； in, and They represent from arrive The actual and nominal models of the open-loop transfer function are represented by the product of the transfer functions of the system's forward path: ； in, For a feedback controller, choose a PI type: , Indicates when the load comes into contact with the environment arrive The nominal model open-loop transfer function; in addition, Defined as the closed-loop transfer function with a nominal model: ； The above equation represents the complementary sensitivity function, which will be used here. Redefining According to the small gain theorem, the robust stability condition of the system is expressed by the following equation: ； Closed-loop control system must meet To achieve robust stability.

Images