Robotic arm control method, device, electronic equipment and medium

Abstract

The present application provides a robotic arm control method, apparatus, electronic device, and medium, relating to the field of robotic arm control. The method is applied to a first follower and includes: estimating the motion state of a leader in the robotic arm at a first moment to obtain a first motion state estimate, wherein the robotic arm includes a leader, a first follower, and at least one second follower; obtaining a second motion state estimate sent by the second follower, wherein the second motion state estimate is obtained by the second follower estimating the motion state of the leader at the first moment; fusing the first motion state estimate with the second motion state estimate to obtain a target motion state estimate; determining a target control signal based on the target motion state estimate, and adjusting the motion state of the first follower based on the target control signal. This ensures consistency in the actions of the multiple components of the robotic arm.

Description

Technical Field

[0001] The present invention relates to the field of robotic arm control, and in particular to a robotic arm control method, device, electronic equipment and medium. Background Art

[0002] Currently, the design and development of robotic arm systems that can adapt to confined industrial operating environments has become a key research topic in related industries. For example, a rope-driven snake-like robotic arm, with its multiple degrees of freedom and flexible structure, can not only nimbly avoid complex obstacles but also reach areas that are inaccessible to traditional robotic arms, fundamentally solving the technical bottleneck of working in confined spaces. The robotic arm consists of a leader and multiple followers, each of which controls and adjusts its own behavior based on the leader's state.

[0003] However, this type of robotic arm contains complex disturbances from various aspects. These complex disturbances will increase the control errors between the multiple components of the robotic arm (multiple components, namely leaders and multiple followers), resulting in inconsistent actions between the multiple components of the robotic arm. Summary of the Invention

[0004] The purpose of the embodiments of the present application is to provide a method, device, electronic device and medium for controlling a robotic arm, so that the actions of multiple components of the robotic arm are consistent.

[0005] In a first aspect, an embodiment of the present application provides a robotic arm control method, applied to a first follower, the method comprising:

[0006] estimating a motion state of a leader in a robotic arm at a first moment to obtain a first motion state estimation value, the robotic arm comprising a leader, the first follower, and at least one second follower;

[0007] Obtaining a second motion state estimation value sent by the second follower, wherein the second motion state estimation value is obtained by the second follower estimating the motion state of the leader at the first moment;

[0008] fusing the first motion state estimation value and the second motion state estimation value to obtain a target motion state estimation value;

[0009] A target control signal is determined according to the target motion state estimation value, and the motion state of the first follower is adjusted according to the target control signal.

[0010] In some embodiments, after adjusting the motion state of the first follower according to the target control signal, the method further includes:

[0011] Acquiring a motion state measurement value, where the motion state measurement value is obtained by measuring the motion state of the first follower using a sensor configured in the first follower;

[0012] Calculating according to the motion state measurement value and the first motion state estimation value to obtain a following error of the first follower;

[0013] determining a compensation control signal according to the following error;

[0014] The motion state of the first follower is controlled using the compensation control signal.

[0015] In some embodiments, determining a compensation control signal according to the following error includes:

[0016] determining the compensation control signal according to the following error and an unknown dynamic compensation amount, wherein the unknown dynamic compensation amount is obtained by inputting the motion state measurement value into a preset neural network model;

[0017] According to the motion error, each parameter in the preset neural network model is updated, wherein the motion error is the difference between the motion state measurement value and the motion state reference value, and the motion state reference value is obtained by the leader measuring the motion state of the leader through a sensor configured in the leader.

[0018] In some embodiments, the second motion state estimation value is sent by the second follower when a first condition is met, and the first condition is that a first difference between the second motion state estimation value and a historical motion state estimation value is greater than a first threshold, wherein the historical motion state estimation value is the motion state estimation value most recently sent by the second follower, and the first threshold is determined based on a first time point, which is the time point when the second follower determines the second motion state estimation value.

[0019] In some embodiments, adjusting the motion state of the first follower according to the target control signal includes:

[0020] Obtaining a second difference value, where the second difference value is a difference between the target control signal and a first control signal, where the first control signal is a control signal most recently used by the first follower;

[0021] When the second difference is greater than a second threshold, the motion state of the first follower is adjusted according to the target control signal, wherein the second threshold is determined according to a second time point, and the second time point is the time point when the first follower determines the target control signal.

[0022] In some embodiments, fusing the first motion state estimate value with the second motion state estimate value to obtain a target motion state estimate value includes:

[0023] Obtaining a third difference value, where the third difference value is a difference between the second motion state estimation value and the first motion state estimation value;

[0024] Multiplying the third difference by a preconfigured coefficient corresponding to the second follower to obtain a target product;

[0025] The target motion state estimate is determined based on the target product.

[0026] In some embodiments, determining the target motion state estimate according to the target product includes:

[0027] updating a transformation matrix according to the target product to obtain a target transformation matrix, wherein the transformation matrix is used to convert an internal state vector into a first-order derivative of the first motion state estimate, and the internal state vector is used to represent an estimation result of the motion state of the leader;

[0028] The target motion state estimate is determined using the target transformation matrix, the internal state vector, and the target product.

[0029] In a second aspect, an embodiment of the present application provides a robotic arm control device, applied to a first follower, the device comprising:

[0030] an observation module, configured to estimate a motion state of a leader in a robotic arm at a first moment to obtain a first motion state estimation value, the robotic arm comprising a leader, the first follower, and at least one second follower;

[0031] An acquisition module, configured to acquire a second motion state estimation value sent by the second follower, wherein the second motion state estimation value is obtained by the second follower estimating the motion state of the leader at a first moment;

[0032] a fusion module, configured to fuse the first motion state estimation value and the second motion state estimation value to obtain a target motion state estimation value;

[0033] A control module is configured to determine a target control signal according to the target motion state estimation value, and adjust the motion state of the first follower according to the target control signal.

[0034] In a third aspect, an embodiment of the present application provides an electronic device, including:

[0035] a memory configured to store instructions; and

[0036] The processor is configured to call the instructions from the memory and to implement the robotic arm control method provided in the first aspect of the embodiment of the present application when executing the instructions.

[0037] In a fourth aspect, an embodiment of the present application provides a machine-readable storage medium, on which instructions are stored. When the instructions are executed by a processor, the processor enables the processor to implement the robotic arm control method described in the first aspect of the embodiment of the present application.

[0038] In an embodiment of the present application, a first follower (each component in the robotic arm can be a first follower) obtains a first motion state estimate by independently estimating the actual motion state of the leader at the first moment; and the first follower also obtains a second motion state estimate sent by at least one second follower (all followers other than the first follower itself are second followers relative to the first follower), and these estimates respectively reflect the local observation results of the same leader state by different followers; then, the processor fuses the first motion state estimate with all received second motion state estimates to obtain a target motion state estimate that fuses the observation results of multiple followers on the leader; finally, the target control signal is calculated based on the target motion state estimate, and the first follower is driven to dynamically adjust its own motion state to track the estimation result represented by the target motion state estimate. Through the above-mentioned distributed multi-source fusion, the first follower combines the first motion state estimation value and at least one second motion state estimation value to generate a target motion state estimation value that can more accurately describe the motion state of the leader, and determines the target control signal based on the target motion state estimation value to achieve consistent motion between the first follower and the leader according to the motion state of the leader according to the target control signal, that is, to achieve consistent motion between multiple followers and the leader (multiple components of the robotic arm). BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 1 is a flow chart of a method for controlling a robotic arm according to an embodiment of the present application;

[0040] Figure 2 This is another flow chart of the robotic arm control method provided in an embodiment of the present application;

[0041] Figure 3 It is a schematic diagram of the experimental results of group 1;

[0042] Figure 4 is a schematic diagram of the experimental results of group 6;

[0043] Figure 5 It is a schematic diagram of tracking error comparison of experimental results;

[0044] Figure 6 Schematic diagram of the structure of the robot arm control device provided in an embodiment of the present application;

[0045] Figure 7 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0046] The following will be combined with the accompanying drawings in the embodiments of the present application to clearly describe the technical solutions in the embodiments of the present application. Obviously, the embodiments described are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field are within the scope of protection of this application.

[0047] The terms "first," "second," and the like in the specification and claims of this application are used to distinguish similar objects, and are not used to describe a specific order or precedence. It should be understood that the terms used in this manner are interchangeable where appropriate, so that the embodiments of this application can be implemented in an order other than that illustrated or described herein, and that the objects distinguished by "first," "second," and the like are generally of the same type, and do not limit the number of objects; for example, the first object can be one or more. In addition, the term "and / or" in the specification and claims represents at least one of the connected objects, and the character " / " generally indicates that the objects associated with each other are in an "or" relationship.

[0048] Specialized robots are now widely used in diverse fields, including underwater, security, aerospace, and petrochemical engineering, as well as in hazardous environments. However, for some operating environments characterized by confined spaces, complex obstacles, and enclosed structures, effective automated maintenance equipment is still lacking, forcing skilled manual labor to perform tasks with low efficiency. For example, the complex piping within aircraft fuel tanks makes conventional automated maintenance equipment difficult to complete. The closely spaced blades of aircraft engines create confined spaces, making it difficult for conventional robotic arms to reach them. Furthermore, operating in confined spaces in high-risk environments, such as those with hazardous gases or strong radiation, requires additional manpower, material resources, and effort. Therefore, the design and development of robotic arms capable of adapting to confined industrial operating environments has become a key research topic in these industries. This led to the design and application of the rope-driven snake-shaped robotic arm. Leveraging its multiple degrees of freedom and flexible structure, it can not only nimbly navigate complex obstacles but also reach areas beyond the reach of traditional robotic arms, fundamentally addressing the technical bottleneck of working in confined spaces.

[0049] However, this type of robotic arm contains complex disturbances from various aspects, such as the elasticity of the rope, which will stretch when subjected to force, causing position errors; multiple joints are coupled with each other, and there are complex random disturbances caused by various external factors. These complex disturbances will increase the control error between the multiple components of the robotic arm (multiple components, namely leaders and multiple followers), resulting in inconsistent actions between the multiple components of the robotic arm.

[0050] Based on this, the embodiments of the present application provide a method, device, electronic device and medium for controlling a robotic arm, so that the actions of multiple components of the robotic arm are consistent.

[0051] The following, in conjunction with the accompanying drawings, describes in detail the robot arm control method, device, electronic device and medium provided in the embodiments of the present application through specific embodiments and their application scenarios.

[0052] A robotic arm consists of a leader and multiple (N) followers. The leader and followers can communicate with each other, and the followers can also communicate with each other. The followers use an observer to estimate the leader's position and a controller to control their own motion based on the observed and estimated values, ensuring consistency with the leader's motion. Those skilled in the art will understand that the observer and controller are modular representations of logic code and are not inherently hardware independent. Data processing for both the observer and controller can be performed on the same processor.

[0053] The dynamic model of the robotic arm is expressed by the following formula:

[0054] (1);

[0055] In formula (1), (follower's number), are the follower’s generalized position vector, velocity vector, and acceleration vector, respectively. are the mass inertia matrix, Coriolis force / centrifugal force matrix and gravity matrix of the manipulator, is the drive input received by the follower’s actuator (servo motor or hydraulic cylinder), is an unknown interference term.

[0056] Assume that the leader signal of the robot arm (the signal sent by the leader to the follower) Produced by the following autonomous systems:

[0057] (2);

[0058] in, , are unknown parameters in the system, and is a matrix containing unknown parameters, the matrix is neutrally stable, that is, the matrix All eigenvalues are semisimple and have zero real part. is the output signal of the system (that is, the leader signal mentioned above). It's about polynomials that satisfy For this type of output regulation problem, it is necessary to ensure that the leader system can track the reference signal and suppress external disturbances at the same time.

[0059] The leader signal consists of two parts: the reference signal and external disturbances .assumed and They can be described by the following autonomous systems:

[0060] (3);

[0061] (4);

[0062] in, , , and , are all unknown parameter vectors. Formula (2) is a combination of Formula (3) and Formula (4), where:

[0063]

[0064] In formula (3) and formula (4) and is an unknown parameter in the leader system, so the leader matrix The information in cannot be directly used for the observer design of the follower.

[0065] An autonomous system is one whose operation is independent of external temporal or spatial signals and is instead completely determined by its internal state and laws. Mathematically, an autonomous system can be represented as a set of nonlinear differential equations. In multi-agent consensus, the leader system is often assumed to be autonomous for ease of control and planning. This makes the leader system's motion more controllable and predictable, making it easier to achieve consensus for the robotic arm.

[0066] According to the above conditions, the following distributed control law can be designed:

[0067] (5);

[0068] in, , is the smooth function to be designed, is the internal state vector of the follower’s observer, The virtual control input generated by the follower's observer (the virtual control input represents the input that the follower's observer determines based on its own observation of the leader's motion state and controls its own motion state), The numbers of followers other than the user.

[0069] Therefore, for the manipulator system - Equation (1) and the leader system - Equation (2), given the control law (5), the manipulator system has a closed-loop solution, that is, all signals of the closed-loop system are bounded and the tracking error , tends to a small neighborhood around zero.

[0070] See Figure 1 , is a flow chart of a robotic arm control method provided by an embodiment of the present application, which is applied to the first follower. Figure 1 As shown, the robot arm control method includes the following steps S100 to S400.

[0071] Step S100: estimating a motion state of a leader in a robotic arm at a first moment to obtain a first motion state estimation value, wherein the robotic arm includes a leader, a first follower, and at least one second follower.

[0072] Figure 2 This is another flow chart of the robot arm control method provided in the embodiment of the present application. Please refer to Figure 1-Figure 2 In an embodiment of the present application, the robotic arm control method provided in the first aspect of the embodiment of the present application is applied to a first follower. The first follower is each follower in the robotic arm, and all followers other than the first follower are second followers. The processor of the first follower (hereinafter referred to as the processor) first uses its own observer to observe and estimate the motion state of the leader at the current moment (i.e., the first moment), obtaining a first motion state estimate value. The first motion state estimate value represents the first follower's observation and understanding of the leader's motion state at the first moment.

[0073] Step S200: Acquire a second motion state estimation value sent by a second follower, wherein the second motion state estimation value is obtained by the second follower estimating the motion state of the leader at the first moment.

[0074] In this step, the processor obtains the second motion state estimate sent by the second follower. The second motion state estimate is also obtained by the second follower using its own observer to observe and estimate the motion state of the leader at the current moment (i.e., the first moment). This second motion state estimate represents the second follower's observation and understanding of the leader's motion state at the first moment. If there are multiple second followers, there will also be multiple second motion state estimates.

[0075] Step S300: fusing the first motion state estimation value and the second motion state estimation value to obtain a target motion state estimation value.

[0076] In this step, the processor fuses the first motion state estimation value with all received second motion state estimation values to obtain a target motion state estimation value that fuses the observation results of multiple followers on the leader.

[0077] In some embodiments, fusing the first motion state estimate value with the second motion state estimate value to obtain a target motion state estimate value includes:

[0078] Obtaining a third difference value, where the third difference value is a difference between the second motion state estimation value and the first motion state estimation value;

[0079] Multiplying the third difference by a preconfigured coefficient corresponding to the second follower to obtain a target product;

[0080] The target motion state estimate is determined based on the target product.

[0081] During the fusion process, the processor may first determine a third difference, namely, the difference between the first motion state estimate and the second motion state estimate. This difference represents the difference between the first follower and the second follower's observations of the leader's motion state. The processor then multiplies this third difference by the coefficient corresponding to the second follower to obtain a target product. (If there are multiple second followers, there will also be multiple third differences, with each third difference multiplied by the corresponding second follower coefficient to obtain multiple target products.) If communication between multiple followers is understood as occurring through a communication network topology, this coefficient represents the strength of the communication signal between the first follower and the second follower, i.e., the connection weight in the communication network topology. Finally, the processor uses the obtained target product to determine a target motion state estimate, which is used to subsequently determine the target control signal.

[0082] In some embodiments, determining a target motion state estimate based on the target product includes:

[0083] updating the transformation matrix according to the target product to obtain a target transformation matrix, wherein the transformation matrix is used to convert the internal state vector into a first-order derivative of the first motion state estimate, and the internal state vector is used to represent the estimation result of the motion state of the leader;

[0084] The target motion state estimate is determined using the target transformation matrix, the internal state vector, and the target product.

[0085] In summary of the above two implementations, specifically, the process of using the third difference to fuse the first motion state estimation value and the second motion state estimation value can be obtained from the following formula:

[0086] ；

[0087] in, is the number of the first follower among all followers, is the number of the second follower (subscript The first follower No. second followers), The virtual control input generated by the observer of the first follower is the target motion state estimate, is the target product, is the coefficient corresponding to the second follower;

[0088] It can be expressed by the following formula:

[0089] (7);

[0090] Right now, is a virtual control input, a signal that can track the leader's dynamics and is used to estimate the leader's state .

[0091] is the third difference, is the first motion state estimate, is the second motion state estimate, For the next control cycle The first derivative of It can be obtained from formula (6) Points earned;

[0092] It will be understood by those skilled in the art that the target motion state estimation value The virtual control input generated by the observer is the target control signal generated by the controller of the first follower and acts on the actuator. There is a conversion relationship between them. When performing actual control, the processor Points earned new , reuse Sure ;

[0093] is the triggering moment when the second follower sends the second motion state estimation value to the first follower, is the set of second followers, That is, multiple target products are summed up. is the preset control gain parameter, which represents the target motion state estimation value and The conversion relationship between ; is the current internal state vector of the first follower’s observer, It represents the intermediate state of the observer's internal estimation of the leader's motion state, that is, an estimation result (not the first motion state estimate);

[0094] is the transformation matrix for the “correction gain” in the first follower observer, which transforms the internal state vector Mapped into a virtual input representing the observer's estimate of the leader's state , those skilled in the art will understand that the Represents the new , which fuses multiple target products and injects the fusion results into the existing That is Obtain, and existing That is, the first-order derivative of the current first motion state estimate;

[0095] Ideally, Should converge to the true design matrix (i.e. target transformation matrix); It is an ideal true value gain matrix obtained by technicians through the Sylvester equation during the observer design phase, which is used to ensure that the observation error dynamics meets the preset stability and convergence speed;

[0096] for No. Line (or Column), which is updated by The second follower is in Status difference in dimension and its own internal state component The external accumulation is added; is the internal state vector No. The first component takes the form of a row vector (scalar transpose) and is used to compare with the second follower in the Status difference in dimension Multiply to form a pair No. Update items for rows;

[0097] That is, the processor first uses the target product to transform the matrix Update and then use the updated target transformation matrix And the target product calculates the target motion state estimate, and finally uses the target motion state estimate Determine the target control signal .

[0098] Step S400: determining a target control signal according to the target motion state estimation value, and adjusting the motion state of the first follower according to the target control signal.

[0099] As described in step S300 , the processor finally determines a target control signal using the target motion state estimation value, and adjusts the motion state of the first follower according to the target control signal to approximate the motion state of the leader.

[0100] Through the above steps S100 to S400, the first follower (each component in the robotic arm can be a first follower) obtains a first motion state estimation value by independently estimating the actual motion state of the leader at the first moment; and the first follower also obtains a second motion state estimation value sent by at least one second follower (all followers other than the first follower itself are second followers relative to the first follower), and these estimation values respectively reflect the local observation results of the same leader state by different followers; then, the processor fuses the first motion state estimation value with all received second motion state estimation values to obtain a target motion state estimation value that fuses the observation results of multiple followers on the leader; finally, the target control signal is calculated based on the target motion state estimation value, and the first follower is driven to dynamically adjust its own motion state to track the estimation result represented by the target motion state estimation value. Through the above-mentioned distributed multi-source fusion, the first follower combines the first motion state estimation value and at least one second motion state estimation value to generate a target motion state estimation value that can more accurately describe the motion state of the leader, and determines the target control signal based on the target motion state estimation value to achieve consistent motion between the first follower and the leader according to the motion state of the leader according to the target control signal, that is, to achieve consistent motion between multiple followers and the leader (multiple components of the robotic arm).

[0101] Regarding the above , can be obtained through the following process:

[0102] The virtual control input can be obtained from formula (7): A high-order dynamic model of:

[0103] (8);

[0104] In formula (8), , is a positive integer, … are all scalar coefficients, polynomials All roots λ of are unequal and lie on the imaginary axis.

[0105] In this way, the adjoint matrix can be constructed and the output matrix ,Will The high-order derivatives of are dynamically transformed into a linear state space model, and the adjoint matrix and the output matrix As shown in the following formula:

[0106] (9);

[0107] Define the state vector ,Will and its derivatives are integrated into a state vector to construct the internal state of the observer. Its parameters are adjusted through the adaptive law to achieve real-time estimation of the leader state:

[0108] (10);

[0109] In formula (10), Meet the conditions and .

[0110] like Controllable, Observable, and matrix and The spectra of do not intersect, then there exists a nonlinear matrix Satisfying the Sylvester equation as in formula (10), formula (11) transforms the dynamic characteristics of the leader system into Target dynamics with observer design Connect them.

[0111] (11);

[0112] in, is a Hurwitz matrix, is a column vector, and , .

[0113] Then, the true value gain matrix can be defined as and the internal state of the observer (i.e. the above ), which maps the leader’s dynamics into the observer’s stable dynamic space through matrix transformation.

[0114] (12);

[0115] Furthermore, the following dynamic equation of the observer can be obtained, which provides a stable state evolution model for the observer and enables it to converge to the leader state:

[0116] (13);

[0117] in, , ,matrix ,matrix ,vector .

[0118] The first line in Equation (13) represents the internal state vector (i.e. the above ), the second line in formula (13) represents the internal state vector With virtual input Through to convert.

[0119] Furthermore, define , used as the initial condition for the observer - that is, to ensure that all observers start from zero error, which facilitates stability proof and practical deployment.

[0120] The observer of the first follower can be designed as follows (14), estimating the state of the leader through local communication and adaptive mechanism:

[0121] (14);

[0122] The second line in formula (14) represents the first follower fusing the internal state vector of the second follower with the current virtual input , for its own internal state vector Update and then use Convert to new Then, as represented by the first line of formula (14), for the new Integrate to obtain a new first motion state estimate.

[0123] It can be seen that formula (14) can be understood as the "general skeleton of the observer", while formula (8) is the "neighborhood difference fusion + adaptive gain" to enrich the specific flesh and blood of this skeleton. In other words, the solution represented by formula (8) is a specific implementation form of the solution represented by formula (14).

[0124] In some embodiments, after adjusting the motion state of the first follower according to the target control signal, the method further includes:

[0125] Acquiring a motion state measurement value, where the motion state measurement value is obtained by measuring the motion state of the first follower through a sensor configured in the first follower;

[0126] Calculating the following error of the first follower based on the motion state measurement value and the first motion state estimation value;

[0127] determining a compensation control signal according to the following error;

[0128] The motion state of the first follower is controlled using the compensation control signal.

[0129] In this embodiment, after completing the preliminary motion adjustment of the first follower based on the target control signal, the processor further introduces a layer of closed-loop compensation to eliminate the residual error caused by the estimation bias and execution lag.

[0130] Specifically, the first follower first obtains a motion state measurement value obtained by real-time measurement from a sensor configured in the first follower. The motion state reference value accurately reflects the current joint angle, position and speed information of the first follower; then, the processor compares this motion state measurement value with the first motion state estimation value previously generated locally by the first follower observer, and calculates the following error between the two, that is, the deviation between the predetermined motion state and its actual motion state.

[0131] The technician may first define the error function of the first follower in the processor as:

[0132] (15);

[0133] in, , is the signal sent by the leader to the first follower, which is measured by the leader through the sensor arranged on the leader. It is the motion state reference value, which represents the real motion state of the leader. is the generalized position vector of the first follower, which is obtained by measuring the motion state of the first follower through the sensor configured in the first follower, that is, the motion state measurement value, is the preset error gain;

[0134] The following error represents the deviation between the expected motion state and its actual motion state. According to formula (15), the following error can be obtained as follows: The expression formula is:

[0135] (16);

[0136] Based on the following error, the processor further determines a compensation control signal to replace the executed target control signal (ie, the compensation control signal is a new target control signal). Finally, the first follower uses this set of compensation control signals to drive the actuator, further adjusting its motion state so that its actual movement more accurately follows the predetermined trajectory. This closed-loop compensation mechanism continuously corrects errors even in the presence of model incompleteness, sensor noise, or external disturbances, significantly improving the first follower's tracking consistency and dynamic response accuracy to the leader's motion.

[0137] In some embodiments, determining a compensation control signal based on a following error includes:

[0138] Determining a compensation control signal based on the first intermediate value and an unknown dynamic compensation value, wherein the unknown dynamic compensation value is obtained by inputting a motion state measurement value into a preset neural network model;

[0139] According to the motion error, various parameters in the preset neural network model are updated, wherein the motion error is the difference between the motion state measurement value and the motion state reference value. The motion state reference value is obtained by measuring the motion state of the leader through a sensor configured in the leader.

[0140] In this embodiment, when the following error of the first follower relative to the leader is known, the processor first adds the following error to a set of unknown dynamic compensation quantities output by a preset neural network model to generate a final compensation control signal - where the unknown dynamic compensation quantities are obtained by inputting the motion state measurement values into the neural network model, and the network will output compensation terms used to offset the unknown disturbances and nonlinear dynamics in the system.

[0141] Compensation control signal The generation process of can be expressed by the following formula:

[0142] (17);

[0143] in, is a positive parameter designed by technicians, The unknown dynamic compensation is obtained by the preset neural network model according to the motion state measurement value. The preset neural network function Expressed as:

[0144] (18);

[0145] in, To preset the parameters of the neural network model, is the basis function vector, which is used to convert the input Mapping to a high-dimensional feature space.

[0146] At the same time, in order to keep the neural network model close to the actual system dynamics, the processor also calculates the motion error after each execution - that is, the difference between the motion state measurement value of the first follower and the motion state reference value measured by the leader through its sensor and provided to the first follower in real time ( ) - Update the parameters of the neural network model online to optimize the subsequent unknown dynamic compensation.

[0147] The specific parameter update process can be expressed by the following formula:

[0148] (19);

[0149] is the first-order derivative of the updated parameter, To preset the current parameters of the neural network function, Positive parameters designed by technicians, By motion error Calculated according to formula (15).

[0150] Through this process, "error-gain adjustment-neural network compensation-online adaptation" constitutes a closed-loop compensation control mechanism, which effectively compensates for the deviations caused by model uncertainty and external disturbances, and achieves high-precision correction of the motion state of the first follower.

[0151] In some embodiments, the second motion state estimation value is sent by the second follower when a first condition is met, the first condition being that a first difference between the second motion state estimation value and a historical motion state estimation value is greater than a first threshold, wherein the historical motion state estimation value is the motion state estimation value most recently sent by the second follower, and the first threshold is determined based on a first time point, which is the time point when the second follower determines the second motion state estimation value.

[0152] In this embodiment, to reduce communication bandwidth and energy consumption, the second follower only transmits its estimate of the leader's second motion state to the first follower when there is a significant change in its estimate. Specifically, the second follower compares its current estimate of the leader's motion state with its historical estimate (i.e., the last estimate sent by the second follower). Communication is triggered only when the first difference between the two estimates—that is, the margin of error between the two estimates—exceeds a pre-set first threshold. This first threshold is a threshold established when the second follower completes its second motion state estimate at the first point in time. It serves as a measure of whether the change is significant enough. Only when the estimate fluctuates beyond this threshold is it considered worthy of updating the neighbor's knowledge and transmitting it.

[0153] The communication trigger mechanism ETM-A between the first follower and the second follower can be expressed as the following formula:

[0154] (20);

[0155] Among them, the first difference is defined for , represents the second follower's current second motion state estimate Last trigger time Status (i.e., historical motion state estimates), It is When the second follower sends the second motion state estimation value to the first follower, a data transmission is performed. For this time ( +1 times) The trigger moment when the second follower sends the second motion state estimation value to the first follower is triggered when ETM-A is met and the signal is in the interval Keep constant inside. is a bounded positive exponential decay coefficient, , Is a bounded positive constant of the event trigger threshold and satisfies the condition , , Constitutes the first threshold, where This is the first point in time.

[0156] In this way, through this threshold-based event triggering mechanism, the second follower can significantly reduce the number of unnecessary communications while ensuring that the first follower obtains sufficiently fresh and useful estimation information, effectively reducing network burden and energy consumption.

[0157] In some embodiments, adjusting the motion state of the first follower according to the target control signal includes:

[0158] Obtaining a second difference, where the second difference is a difference between the target control signal and the first control signal, and the first control signal is a control signal most recently used by the first follower;

[0159] When the second difference is greater than a second threshold, the motion state of the first follower is adjusted according to the target control signal, wherein the second threshold is determined according to a second time point, and the second time point is the time point when the first follower determines the target control signal.

[0160] In this embodiment, to further reduce the frequency of control signal updates and smooth actuator response, the processor compares the changes between the newly generated target control signal and its first control signal (i.e., the last control signal actually issued to the actuator and in effect) to calculate a second difference. This second difference quantifies the amplitude difference between the new and old control commands and is compared with a pre-set second threshold at a second point in time (i.e., the moment the new target control signal is determined). If the second difference is less than the threshold, the processor determines that the difference between the new and old control signals is insufficient to significantly affect following performance and delays or skips the update to avoid introducing excessive minor jitter at the actuator. Only when the second difference exceeds the second threshold is the new target control signal issued to the actuator, directly adjusting the motion state of the first follower.

[0161] The above control execution trigger mechanism ETM-B can be expressed as the following formula:

[0162] (twenty one);

[0163] Among them, the second difference is defined for , It is The moment when the first follower sends the control signal to the actuator and it is taking effect, For this time ( +1 time) The moment when the first follower sends the control signal to the actuator, it is triggered when ETM-B is met and the signal is within the interval Keep constant inside. , , is a bounded positive number and satisfies the condition , , Constitute the second threshold, where This is the second time point.

[0164] In some embodiments, during the ETM-B trigger interval Internal, control input Keep the last event time The control input signal value 。

[0165] (twenty two);

[0166] In this way, through this threshold-triggered control update strategy, the first follower can not only respond promptly to large trajectory correction requirements, but also suppress the frequent switching of actuators caused by small jitters, thereby significantly improving the smoothness and energy efficiency of control while ensuring accurate tracking of the leader.

[0167] The following is a robotic arm experiment conducted based on the above multiple implementations to verify the technical effects of the robotic arm control method provided in the embodiments of the present application:

[0168] In this experiment, the three joints of a rope-driven hyper-redundant snake-like manipulator are used as the follower subsystem. The system equation of the leader is as follows:

[0169] (twenty three);

[0170] Its system matrix Set to , , The initial value of .Pick Serves as the leader signal in this experiment.

[0171] The structures of the designed observer and controller and their corresponding event triggering mechanisms are listed in Table 1 below. The relevant parameters are set as follows, where , , The parameters of the radial basis neural network are set as follows: the number of hidden layer nodes , the ideal weight vector , the basis function center and Gaussian kernel width are set as and .

[0172]

[0173] Table 1 Observer and controller structure

[0174] The sampling time of the system is set to 0.01s. Set to a random value between -1 and 1. The experiment is divided into six groups. The first five groups use the control scheme proposed in Chapter 3, and are given different event trigger threshold parameters. The controller of the last group uses an event-triggered PID controller as the experimental control group, and the event trigger threshold parameter setting is the same as the first group. The event trigger threshold parameters used in the experiment are The settings are shown in Table 2. The controller parameters in the first five groups are all set to , , , The PID controller parameters in the sixth group are set to , , .

[0175]

[0176] Table 2 Event trigger threshold parameter table

[0177] Figure 3 This is a schematic diagram of the first group of experimental results. Figure 4 This is a schematic diagram of the sixth set of experimental results, which intuitively shows the tracking performance of the two controllers. Figure 5 The following figure is a schematic diagram comparing the tracking errors of the experimental results. The experimental data of follower 1 in the two groups are plotted, and the root mean square error (RMSE) and mean absolute error (MAE) of the position and velocity of the two are calculated and shown in Table 3. The experimental results show that under this dual-channel event-triggered control structure, the proposed controller can achieve better control performance than the traditional PID controller, and the position and velocity tracking errors are smaller.

[0178]

[0179] Table 3 Root mean square error (RMSE) and mean absolute error (MAE) of system tracking error

[0180] The number of event triggers in the communication channel and controller channel of each group of experiments is recorded in Table 4 below. The experimental results show that the proposed dual-channel event-triggered control strategy can greatly reduce the number of communications in the communication channel and keep the tracking error of the system within a relatively small range.

[0181]

[0182] Table 4 Event triggering times in each channel

[0183] See Figure 6 , is a schematic structural diagram of a robotic arm control device provided in an embodiment of the present application. In a second aspect, an embodiment of the present application provides a robotic arm control device 10, which is applied to a first follower. The device 10 includes:

[0184] An observation module 11 is configured to estimate a motion state of a leader in a robotic arm at a first moment to obtain a first motion state estimation value, wherein the robotic arm includes a leader, a first follower, and at least one second follower;

[0185] An acquisition module 12 is configured to acquire a second motion state estimation value sent by the second follower, wherein the second motion state estimation value is obtained by the second follower estimating the motion state of the leader at the first moment;

[0186] a fusion module 13, configured to fuse the first motion state estimation value and the second motion state estimation value to obtain a target motion state estimation value;

[0187] The control module 14 is configured to determine a target control signal according to the target motion state estimation value, and adjust the motion state of the first follower according to the target control signal.

[0188] The robot arm control device 10 provided in the second aspect of the embodiment of the present application can implement each process implemented in the above method embodiment and achieve the same beneficial effects. To avoid repetition, it will not be repeated here.

[0189] See Figure 7 , is a structural diagram of an electronic device provided in an embodiment of the present application. The third aspect of an embodiment of the present application provides an electronic device 1000, including a processor 1100 and a memory 1200. The memory 1200 stores machine-executable instructions that can be executed by the processor 1100. The processor 1100 can execute the machine-executable instructions to implement the above-mentioned robotic arm control method.

[0190] A fourth aspect of an embodiment of the present application provides a machine-readable storage medium, on which instructions are stored. When the instructions are executed by a processor, the processor implements the above-mentioned robotic arm control method.

[0191] In some embodiments, the embodiments of the present application further provide a computer program product, including a computer program, which implements the robotic arm control method according to the above embodiment when executed by a processor.

[0192] Those skilled in the art will appreciate that the embodiments of the present application can be provided as methods, systems, or computer program products. Therefore, the present application can adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application can adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code.

[0193] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the steps in the process. Figure 1 a process or multiple processes and / or boxes Figure 1 These computer program instructions can also be stored in a computer-readable memory that can guide a computer or other programmable data processing device to work in a specific way, so that the instructions stored in the computer-readable memory produce a product including the instruction device, which implements the function specified in the process. Figure 1 a process or multiple processes and / or boxes Figure 1 These computer program instructions can also be loaded onto a computer or other programmable data processing device, so that a series of operation steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process in the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0194] In a typical configuration, a computing device includes one or more processors (CPUs), input / output interfaces, network interfaces, and memory.

[0195] The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. The memory is an example of a computer-readable medium.

[0196] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can be implemented using any method or technology to store information. Information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassettes, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer-readable media does not include temporary computer-readable media such as modulated data signals and carrier waves.

[0197] It should also be noted that the terms "comprises," "includes," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a process, method, commodity, or apparatus that includes a series of elements includes not only those elements but also other elements not explicitly listed, or includes elements inherent to such process, method, commodity, or apparatus. In the absence of further limitations, an element defined by the phrase "comprises a ..." does not exclude the presence of other identical elements in the process, method, commodity, or apparatus that includes the element.

[0198] The above are merely embodiments of the present application and are not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present application should all be included within the scope of the claims of the present application.

[0199] In addition, the various embodiments of the present invention may be arbitrarily combined, and as long as they do not violate the concept of the present invention, they should also be regarded as the contents disclosed by the present invention.

Claims

1. A method for controlling a robotic arm, characterized in that: Applied to the first follower, the method includes: estimating a motion state of a leader in a robotic arm at a first moment to obtain a first motion state estimation value, the robotic arm comprising a leader, the first follower, and at least one second follower; Obtaining a second motion state estimation value sent by the second follower, wherein the second motion state estimation value is obtained by the second follower estimating the motion state of the leader at a first moment, and the second motion state estimation value is sent by the second follower if a first condition is satisfied, the first condition being that a first difference between the second motion state estimation value and a historical motion state estimation value is greater than a first threshold, wherein the historical motion state estimation value is a motion state estimation value most recently sent by the second follower, and the first threshold is determined based on a first time point, which is the time point at which the second follower determines the second motion state estimation value; fusing the first motion state estimation value and the second motion state estimation value to obtain a target motion state estimation value; A target control signal is determined according to the target motion state estimation value, and the motion state of the first follower is adjusted according to the target control signal.

2. The method according to claim 1, characterized in that After adjusting the motion state of the first follower according to the target control signal, the method further includes: Acquiring a motion state measurement value, where the motion state measurement value is obtained by measuring the motion state of the first follower using a sensor configured in the first follower; Calculating according to the motion state measurement value and the first motion state estimation value to obtain a following error of the first follower; determining a compensation control signal according to the following error; The motion state of the first follower is controlled using the compensation control signal.

3. The method according to claim 2, characterized in that The determining of the compensation control signal according to the following error comprises: determining the compensation control signal according to the following error and an unknown dynamic compensation amount, wherein the unknown dynamic compensation amount is obtained by inputting the motion state measurement value into a preset neural network model; According to the motion error, each parameter in the preset neural network model is updated, wherein the motion error is the difference between the motion state measurement value and the motion state reference value, and the motion state reference value is obtained by the leader measuring the motion state of the leader through a sensor configured in the leader.

4. The method according to claim 1, wherein The adjusting the motion state of the first follower according to the target control signal includes: Obtaining a second difference value, where the second difference value is a difference between the target control signal and a first control signal, where the first control signal is a control signal most recently used by the first follower; When the second difference is greater than a second threshold, the motion state of the first follower is adjusted according to the target control signal, wherein the second threshold is determined according to a second time point, and the second time point is the time point when the first follower determines the target control signal.

5. The method according to claim 1, wherein The fusing the first motion state estimation value and the second motion state estimation value to obtain a target motion state estimation value includes: Obtaining a third difference value, where the third difference value is a difference between the second motion state estimation value and the first motion state estimation value; Multiplying the third difference by a preconfigured coefficient corresponding to the second follower to obtain a target product; The target motion state estimate is determined based on the target product.

6. The method according to claim 5, characterized in that Determining the target motion state estimation value according to the target product includes: updating a transformation matrix according to the target product to obtain a target transformation matrix, wherein the transformation matrix is used to convert an internal state vector into a first-order derivative of the first motion state estimate, and the internal state vector is used to represent an estimation result of the motion state of the leader; The target motion state estimate is determined using the target transformation matrix, the internal state vector, and the target product.

7. A robotic arm control device, characterized in that: Applied to the first follower, the device includes: an observation module, configured to estimate a motion state of a leader in a robotic arm at a first moment to obtain a first motion state estimation value, the robotic arm comprising a leader, the first follower, and at least one second follower; an acquisition module, configured to acquire a second motion state estimation value sent by the second follower, wherein the second motion state estimation value is obtained by the second follower estimating the motion state of the leader at a first moment, and the second motion state estimation value is sent by the second follower when a first condition is satisfied, the first condition being that a first difference between the second motion state estimation value and a historical motion state estimation value is greater than a first threshold, wherein the historical motion state estimation value is the motion state estimation value most recently sent by the second follower, and the first threshold is determined based on a first time point, which is the time point at which the second follower determines the second motion state estimation value; a fusion module, configured to fuse the first motion state estimation value and the second motion state estimation value to obtain a target motion state estimation value; A control module is configured to determine a target control signal according to the target motion state estimation value, and adjust the motion state of the first follower according to the target control signal.

8. An electronic device, characterized in that: The method comprises a processor and a memory, wherein the memory stores a program or instruction that can be run on the processor, and when the program or instruction is executed by the processor, the method for controlling the robotic arm according to any one of claims 1 to 6 is implemented.

9. A machine-readable storage medium, characterized in that The machine-readable storage medium stores instructions, which, when executed by a processor, enable the processor to implement the robotic arm control method according to any one of claims 1 to 6.

Images