Layered architecture networked robot time-varying optimization formation control method

Abstract

The invention discloses a time-varying optimization formation control method for networked robots with a layered architecture, and the method comprises the steps: collecting the pose and kinetic parameter information of the networked robots, carrying out the equation conversion based on kinematics and kinetic models, and constructing a chain control form capable of describing the state change; track correction is realized by setting a cost function and optimizing a signal generator; a nonlinear filtering function and a virtual controller are introduced to suppress position and speed errors; a radial basis function neural network is adopted to carry out approximation on an unknown nonlinear dynamic state, an adaptive control law is constructed in combination with error feedback, and time-varying optimization of the networked robot formation is realized. According to the method, the convergence speed, stability and robustness of the system can be effectively improved in dynamic and uncertain environments, it is guaranteed that the robot group achieves accurate formation control in complex tasks, and the method has high engineering application value.

Description

Technical Field

[0001] This invention relates to the field of networked robot control technology, and in particular to a time-varying optimized formation control method for networked robots with a hierarchical architecture. Background Technology

[0002] With the deep integration of network technology and intelligent equipment, networked robots are increasingly widely used in military, commercial, and civilian fields. For example, in the military, they can be used for monitoring and rescue missions; in the commercial sector, they can optimize energy allocation; and in the civilian sector, they can perform tasks such as agricultural coverage. This trend places higher demands on the collaborative control capabilities of networked robots. While existing collaborative control schemes can solve basic collaboration problems, they struggle to achieve optimal performance indicators. Furthermore, the movement characteristics of targets, the dynamic changes in the environment, and the time-varying requirements of tasks in real-world scenarios pose further stringent challenges to networked robots—they not only need to maintain a preset formation configuration but also achieve optimal real-time tracking of dynamic targets through information interaction within their local neighborhood. Therefore, balancing the combined requirements of configuration stability and optimal tracking has become a core challenge in distributed formation tracking, and exploring time-varying optimized formation control strategies suitable for networked robots has become a critical issue that urgently needs to be addressed.

[0003] In the field of distributed cooperative control, existing research has yielded numerous achievements in addressing optimization problems. For example, scholars have conducted research on optimal formation problems for Eulerian-Lagrange systems and unmanned surface vessels, achieving distributed formation and optimization objectives for each node. However, existing distributed optimization schemes are mostly designed for time-invariant scenarios, making them difficult to adapt to formation tracking tasks with moving targets. In practical applications, the time-varying characteristics of local cost functions are widespread: for example, in power system dispatching, time-of-use electricity price fluctuations directly alter the form of the power generation cost function; in intelligent transportation, the energy consumption optimization objective of vehicle clusters is updated in real time according to road conditions. Therefore, exploring time-varying optimization problems has become a key direction for solving the dynamic cooperative control of networked robots, and it is also the technical bottleneck that this invention aims to overcome.

[0004] Therefore, it is evident that traditional single control methods have significant limitations when facing the combined requirements of time-varying optimization and time-bound formation control in networked robot systems. They struggle to handle dynamically changing optimization objectives and cannot guarantee high-precision formation completion within a preset time, making it difficult to achieve coordinated unity between the two. To address this, this invention innovatively introduces a hierarchical architecture, systematically decoupling the two core tasks of optimization and control to construct a hierarchical collaborative control system: the first layer (optimization layer) focuses on the dynamic optimization of time-varying trajectories, generating reference trajectories that meet the global optimization objective in real time through a decision-making mechanism adapted to time-varying characteristics; the second layer (control layer) focuses on formation control execution within a specified time, ensuring that the robot swarm converges to the target configuration within a preset time and that the tracking accuracy meets task requirements by constructing a predefined time controller. This hierarchical decoupling design not only overcomes the performance bottlenecks of traditional methods in complex scenarios but also achieves efficient coordination between optimization and control, providing a new technical path for solving the complex control challenges of networked robot systems.

[0005] To address these issues, we designed a hierarchical control method for networked robots that combines a time-varying trajectory optimization signal generator and a predefined time-grouping controller. Summary of the Invention

[0006] The purpose of this invention is to solve the technical problems mentioned in the background section and to provide a hierarchical networked robot time-varying optimization formation control method, comprising the following steps:

[0007] S1. Collect parameter information of the networked robot and assign initial state values ​​to the networked robot;

[0008] S2. Based on the kinematic and dynamic equations related to pose, transform them to obtain the chain control equations that can describe the changes in the robot's state.

[0009] S3. Determine the cost function and construct an optimized signal generator;

[0010] S4. Construct a nonlinear filtering function and a virtual controller;

[0011] S5. Construct a controller based on error relationships and neural networks to achieve time-varying optimization control of networked robot formations.

[0012] As a preferred technical solution of the present invention, step S1 specifically includes collecting state variables such as position, velocity, heading angle and its derivative of the networked robot in a two-dimensional plane, and constructing kinematic and dynamic models based on this information, wherein position represents the position in the geographic coordinate system, heading angle is used to characterize orientation, and velocity and angular velocity are used to describe the changes in linear velocity and angular velocity, respectively.

[0013] As a preferred technical solution of the present invention, step S2 specifically includes transforming the robot dynamics equation to obtain a chain expression that can simultaneously reflect the relationship between the rate of change of position, velocity, angular velocity and control input, thereby providing a basis for subsequent formation optimization control.

[0014] As a preferred technical solution of the present invention, step S3 specifically includes: determining a cost function to characterize the deviation between the robot's current position and the desired target position; and designing and optimizing a signal generator based on the cost function to dynamically correct the robot's running trajectory so that it gradually approaches the optimal trajectory.

[0015] As a preferred technical solution of the present invention, step S4 specifically includes constructing a nonlinear filtering function, which is divided into different stages in time. By using a segmented design, the control signal is smoothly adjusted, thereby avoiding control jitter caused by sudden changes and realizing continuous optimization of the formation.

[0016] As a preferred technical solution of the present invention, step S4 further includes, in order to suppress formation error, constructing a new error variable by combining the position error with a nonlinear transformation function, and then constructing a virtual controller based on the new error variable to achieve effective tracking control of the position error.

[0017] As a preferred embodiment of the present invention, step S5 specifically includes: firstly, constructing a first-order filter to address the differential explosion problem caused by derivatives in the backstepping controller design; and then constructing a speed tracking error based on the filter error, providing a foundation for the subsequent controller design.

[0018] As a preferred technical solution of the present invention, step S5 further includes using an RBF neural network to approximate the unknown nonlinear dynamics, wherein the input of the neural network is the state quantity in the formation control, and the output is an estimate of the unknown dynamic term, thereby compensating for the modeling error and the influence of external disturbances.

[0019] As a preferred technical solution of the present invention, step S5 further includes constructing a control law based on the neural network. The control law is jointly acted upon by the error quantity, the neural network estimate, and the virtual control input to ensure that the robot can maintain stable operation during the formation process and meet the desired trajectory constraints.

[0020] As a preferred technical solution of the present invention, step S5 further includes designing an adaptive law for the neural network, and achieving continuous approximation and correction of nonlinear dynamics by dynamically adjusting the weight parameters, thereby ensuring the robustness and convergence of the system in long-term operation.

[0021] Beneficial effects:

[0022] This application proposes a hierarchical design method for the formation control problem of networked robots, achieving the following technical effects:

[0023] The hierarchical networked time-varying optimization formation control method for robots proposed in this invention achieves dynamic adjustment of the robot formation state through a complete chain control framework of "parameter acquisition—dynamic transformation—optimization signal generation—nonlinear filtering—neural network control law". Compared with traditional methods, this method decouples time-varying optimization and formation control. The optimal signal generator mainly realizes the estimation of time-varying trajectories and achieves consistency. The control layer is designed to fully consider the combination of kinematic and dynamic characteristics, which can reflect the time-varying characteristics of robot position, velocity and heading, and closely link control input and error feedback, thereby ensuring the continuity and real-time performance of the formation control process.

[0024] This invention effectively eliminates positional and velocity errors during robot formation by constructing a cost function, optimizing the signal generator, and using a virtual controller. It also utilizes a nonlinear filtering function to smooth the control signal, avoiding system jitter caused by sudden changes. Furthermore, by introducing error constraint variables and virtual control inputs, the system's adaptability to transient performance is further enhanced, enabling the robot swarm to maintain stable formation and good convergence performance even in complex dynamic environments.

[0025] This invention utilizes radial basis function neural networks to approximate unknown nonlinear dynamics in real time, and designs an adaptive law to dynamically adjust the neural network weights, effectively solving the problem of difficult accurate modeling in traditional control methods. This method not only maintains high control accuracy under uncertainties and external disturbances, but also possesses good adaptability and scalability, making it widely applicable in scenarios such as UAV swarm control, unmanned vehicle platooning, and collaborative robot systems, demonstrating significant engineering application value and promotional significance. Attached Figure Description

[0026] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below.

[0027] Figure 1 This is a structural diagram of the present invention;

[0028] Figure 2 This is a 3D trajectory diagram of the networked robot of this invention;

[0029] Figure 3 This is a formation trajectory diagram of the networked 2D planar robot of this invention;

[0030] Figure 4This invention relates to networked robot control input. The image;

[0031] Figure 5 This invention relates to networked robot control input. The image. Detailed Implementation

[0032] To make the objectives, technical solutions, and advantages of this invention clearer, the following description is provided in conjunction with embodiments and appendices. Figures 1-5 The present invention will be further described in detail below. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0033] Example 1:

[0034] A time-varying optimization formation control method for networked robots with a hierarchical architecture, characterized by the following steps:

[0035] Step S1: Collect the parameter information of the networked robot and assign initial values ​​to the state of the networked robot. The specific operations of Step S1 are as follows: Collect the parameter information of the networked robot; its kinematic and dynamic equations are as follows:

[0036]

[0037] in These represent the x-axis position, y-axis position, and heading angle of the i-th networked robot in the geodetic coordinate system, respectively. Let represent the linear velocity and angular velocity of the i-th networked robot, respectively; Let x and y represent the derivatives of the x-axis position, y-axis position, and heading angle of the i-th networked robot, respectively. Let represent the mass constants of the linear velocity and angular velocity of the i-th networked robot, respectively; Let represent the control torques for the linear velocity and angular velocity of the i-th networked robot, respectively.

[0038] Step S2: Transform the kinematic and dynamic equations of the networked robot using the hand-position method to obtain the transformed chain control equations; Step two is specifically operated as follows:

[0039] The kinematic and dynamic equations of the networked robot are transformed using the hand-position method, and the transformation formula is as follows:

[0040]

[0041] in, The derivative represents the position information of the i-th networked robot after transformation. This represents the speed information of the i-th networked robot after transformation; The derivative representing the converted velocity information of the i-th networked robot; Represents unknown dynamic information after transformation; This represents the control input for the i-th networked robot after the transformation.

[0042] Step S3: Determine the cost function and construct the optimal signal formation generator; the specific operations of step three are as follows:

[0043] Step 3.1: Determine the cost function as follows:

[0044]

[0045] Where t represents time. An estimated value representing the position information of the i-th networked robot; This represents the optimal solution.

[0046] Step 3.2, construct the following optimal signal generator:

[0047]

[0048] in, An estimated value representing the position information of the i-th networked robot; Functions representing gradient dependence; A collection representing network robots; An estimated value representing the position information of the j-th networked robot; Represents the parameters of the design; Represents the inverse of the Hessian matrix; The first derivative of the cost function; The partial derivatives of the first derivative of the valence function; sgn represents the sign function; The transformed value representing the symbolic function; It is a parameter for consistency;

[0049] Step S4 involves constructing the nonlinear transformation function and the formation error of the control layer. The specific operations are as follows:

[0050] Step 4.1: Construct the following nonlinear transformation function:

[0051]

[0052] Where t is time; These are time parameters designed by humans; These are custom parameters. This function is primarily used to convert errors, thereby achieving formation control within a specified time.

[0053] Step 4.2: To achieve the desired control effect, construct the following error transformation.

[0054]

[0055] in, This represents the position tracking error of the i-th networked robot; Represents the formation distance of the i-th networked robot; It is the first error after the i-th networked robot is transformed; This represents the second tracking error of the i-th networked robot; The output of the filter representing the i-th networked robot; It is the filtering error of the i-th networked robot; This represents the virtual control input for the i-th networked robot.

[0056] The virtual control input is constructed as follows:

[0057]

[0058] in Parameters representing virtual control inputs; This represents the position tracking error of the i-th networked robot; The derivative of the formation distance of the i-th networked robot; The derivative representing the estimated position information of the i-th networked robot; It represents a known function.

[0059] Step S5: Construct a controller based on the error relationship and the neural network. The specific operation is as follows:

[0060] Step 5.1: First, construct the following first-order linear filter:

[0061]

[0062] in Parameters representing filter design; The derivative of the output of the i-th networked robot filter; This represents the output of the i-th networked robot filter; This represents the virtual control input for the i-th networked robot.

[0063] Step 5.2: Select an RBF neural network to approximate the unknown robot dynamics; the specific form is as follows:

[0064]

[0065] Where F represents the unknown nonlinear dynamics. This represents the weights of the neural network for the i-th networked robot. T represents the transpose. X represents the radial basis function; X represents the input of the radial basis function. This represents the approximation error.

[0066] Step 5.3: Construct the following control input, in the following specific form:

[0067]

[0068] in, Parameters representing control inputs; This represents the second tracking error of the i-th networked robot; The estimated weights of the neural network representing the i-th networked robot; The radial basis function represents the i-th networked robot; This represents the derivative of the output of the i-th networked robot filter.

[0069] Step 5.4: Construct the following adaptive law: Its specific form is as follows:

[0070]

[0071] in The derivative of the estimated weight of the i-th networked robot; The first design parameter representing the adaptive law; The radial basis function represents the i-th networked robot; The first design parameter representing the adaptive law; This represents the estimated weight of the i-th networked robot.

[0072] Based on the above control method using an optimal time-varying trajectory signal generator and performance constraint control, the intelligent agent cluster can calculate the desired velocity and direction of motion based on its own position, current local observation information, and global observation information, and transmit them to the actuator for motion control, thereby realizing the task of exploring the optimal trajectory and formation control.

[0073] Example 2: In this example, the hierarchical networked robot time-varying optimization formation control method is applied to the unmanned vehicle cluster collaborative inspection task in two-dimensional space. This example is optimized for inspection requirements in complex spatial environments.

[0074] First, in step S1, the pose information of the autonomous vehicle is collected, including its position coordinates.

[0075] (Velocity vector, heading angle, and other parameters are assigned initial state values ​​to each autonomous vehicle. Based on this information, a kinematic and dynamic model is constructed, where the state variables include linear velocity and angular velocity.)

[0076] In step S2, the dynamic equations of the UAV are transformed to obtain a chain-like control equation that simultaneously reflects the relationship between the rate of change of position, velocity, angular velocity, and control input. This chain-like equation provides the mathematical foundation for achieving precise trajectory tracking and formation keeping in subsequent unmanned missions.

[0077] In step S3, a cost function is designed based on the spatial distribution of the inspection target points to characterize the deviation between the current position of the unmanned vehicle and the target spatial points, and a path constraint term is introduced to ensure that the UAVs avoid obstacles during formation flight. The optimized signal generator will correct the trajectory in real time according to the time-varying inspection task (such as dynamic updates of target points) during the calculation process.

[0078] In step S4, a nonlinear filtering function is used to smooth the control signal, avoiding sudden changes in commands caused by path switching. Simultaneously, a virtual controller is constructed to eliminate tracking errors, thereby maintaining formation stability.

[0079] In step S5, an RBF neural network is used to approximate complex airflow disturbances and aerodynamic nonlinearities. The neural network input is the current state variables of the unmanned vehicle (position, velocity, attitude), and the output is an estimate of the unknown disturbance. An adaptive control law is constructed by combining error feedback to achieve dynamic compensation for external disturbances. By designing an adaptive law to dynamically adjust the neural network weight parameters, the robustness of the system during long-term flight is further improved.

[0080] Using the method described in this embodiment, unmanned vehicle swarms can maintain stable formation and achieve high-precision inspection tasks in complex environments such as high-rise building clusters and power transmission lines. Compared with traditional control methods, the method in this embodiment not only maintains fast convergence speed and high stability in dynamic environments, but also exhibits strong robustness and adaptability when facing wind field disturbances and time-varying targets.

[0081] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A layered architecture networked robot time-varying optimization formation control method, characterized in that, The method comprises the following steps: S1, collecting parameter information of the networked robot and assigning an initial state value to the networked robot; S2, transforming according to the kinematics and dynamics equations related to the pose to obtain a chain control equation describing the state change of the robot; S3, determining a cost function and constructing an optimization signal generator; S4, constructing a nonlinear filtering function and constructing a virtual controller; S5, constructing a controller according to the error relationship and the neural network to realize time-varying optimization control of the networked robot formation.

2. The layered architecture networked robot time-varying optimization formation control method according to claim 1, characterized in that: The step S1 specifically comprises collecting the position, velocity, heading angle and derivative state quantity of the networked robot in a two-dimensional plane, and constructing a kinematics and dynamics model according to the information, wherein the position represents the position in the geographical coordinate system, the heading angle is used to represent the direction, and the velocity and angular velocity are used to describe the changes of linear velocity and angular velocity.

3. The layered architecture networked robot time-varying optimization formation control method of claim 1, wherein: The step S2 specifically comprises transforming the robot dynamics equation to obtain a chain expression reflecting the relationship between the position change rate, velocity, angular velocity and control input, thereby providing a basis for subsequent formation optimization control.

4. The layered architecture networked robot time-varying optimization formation control method of claim 1, wherein: The step S3 specifically comprises determining a cost function for describing the deviation between the current position of the robot and the desired target position, and designing an optimization signal generator based on the cost function to dynamically correct the robot running trajectory so that it gradually approaches the optimal trajectory.

5. The layered architecture networked robot time-varying optimization formation control method of claim 1, wherein: The step S4 specifically comprises constructing a new error variable by combining the position error with a nonlinear transformation function, and then constructing a virtual controller according to the new error variable to realize effective tracking control of the position error.

6. The layered architecture networked robot time-varying optimization formation control method of claim 1, wherein: The step S4 further comprises constructing a virtual controller to suppress the formation error, and defining a virtual control input based on the position error, velocity error and filtering error to eliminate the tracking deviation of position and velocity and improve the stability of the system.

7. The layered architecture networked robot time-varying optimization formation control method of claim 1, wherein: The step S5 specifically comprises processing the differential explosion problem caused by the derivative of the controller designed by the backstepping method. The tracking error of the velocity is constructed based on the filtering error to provide a basis for the design of the subsequent controller.

8. The layered architecture networked robot time-varying optimization formation control method of claim 1, wherein: The step S5 further comprises using an RBF neural network to approximate the unknown nonlinear dynamics, wherein the input of the neural network is the state quantity in the formation control, and the output is the estimation of the unknown dynamic term, thereby compensating for the influence of modeling errors and external disturbances.

9. The layered architecture networked robot time-varying optimization formation control method of claim 1, wherein: The step S5 further comprises constructing a control law based on the neural network, wherein the control law is jointly acted on by the error quantity, the neural network estimation quantity and the virtual control input, so as to ensure that the robot can maintain stable operation during the formation process and meet the expected trajectory constraint.

10. The layered architecture networked robot time-varying optimization formation control method of claim 1, wherein: The step S5 further comprises designing an adaptive law of the neural network to realize continuous approximation and correction of the nonlinear dynamics by dynamically adjusting the weight parameters, thereby ensuring the robustness and convergence of the system in long-term operation.

Images