Cable inspection robot attitude control method based on fusion of MPC and ADRC

Abstract

The invention discloses a cable inspection robot attitude control method based on MPC and ADRC fusion, and the method comprises the steps: firstly, building a model prediction control (MPC) model, carrying out the prediction of the future state of a robot, solving an optimal control problem, and obtaining initial control input for achieving the internal dynamic compensation; secondly, an extended state observer (ESO) in an auto-disturbance rejection control (ADRC) model is adopted, and external disturbance is estimated in real time and used for compensating control input; thirdly, a fuzzy self-adaptive module based on fuzzy PID is introduced, the attitude error, the attitude error change rate and regeneration disturbance intensity generated based on estimation disturbance are collected in real time, the MPC weight and the ESO gain are dynamically adjusted, and the response speed and robustness of the system under variable working conditions are improved. According to the hierarchical control scheme, the defect that an existing single control method is difficult to keep the optimal control effect under variable working conditions is effectively overcome, and technical guarantee is provided for stable operation of the cable inspection robot in a high-risk and high-complexity environment.

Description

Technical Field

[0001] The present invention belongs to the field of robot control, and in particular relates to a cable inspection robot posture control method based on the fusion of MPC and ADRC. Background Art

[0002] In recent years, with the increasing demand for intelligent inspection of cables, such as transmission lines, cable inspection robots, as a new type of aerial work equipment, have attracted considerable attention for their posture control technology. Currently available literature primarily employs single control methods, such as fuzzy PID, Active Disturbance Rejection Control (ADRC), or Model Predictive Control (MPC), to manage robot posture changes during navigation and obstacle avoidance. Traditional fuzzy PID controllers are widely used due to their simple structure and easy debugging. However, they often suffer from slow response speed and limited adjustment accuracy in dynamic robot processes characterized by multiple degrees of freedom, nonlinearity, and strong coupling. While ADRC offers strong disturbance rejection capabilities, it lacks the ability to predict internal dynamic changes, such as posture fluctuations caused by leg lifting, speed changes, and obstacle traversal. While MPC methods can predict and globally optimize future states, their fixed parameter design struggles to adapt to transient disturbances such as wind and cable sway in complex environments with stringent real-time requirements, reducing overall control robustness.

[0003] In summary, existing technologies mostly focus on a single control strategy, fail to fully take into account the dual impact of internal dynamics and external disturbances, and lack environmental adaptability, making it difficult to maintain optimal control effects under changing working conditions. Summary of the Invention

[0004] To address the shortcomings of existing technologies, the present invention provides a posture control method for a cable inspection robot based on the fusion of MPC and ADRC. This method aims to address the problem of posture instability caused by the robot's internal posture fluctuations during actions such as obstacle crossing and leg lifting, as well as external disturbances such as high-altitude wind and cable sway. This control method utilizes a model predictive control (MPC) module to dynamically predict and compensate for the robot's internal posture fluctuations, and an active disturbance rejection control (ADRC) module to estimate and compensate for external disturbances in real time. Furthermore, a fuzzy adaptive module is introduced to adjust the outputs of the (MPC) and (ADRC) modules, significantly improving the system's response speed and robustness under variable operating conditions.

[0005] To achieve the above object, the present invention provides the following technical solutions:

[0006] A cable inspection robot posture control method based on the fusion of MPC and ADRC is characterized by comprising the following steps:

[0007] Step a: Build a control model, including:

[0008] Based on the Newton-Euler method, a state-space model is established for the robot system state x(k), the system control input u(k), and the external disturbance intensity d(k); based on the state-space model, an MPC model and an MPC optimization problem are constructed; based on the state-space model, an ADRC model and its extended state observer (ESO) are constructed; based on the fuzzy PID, a fuzzy adaptive module and its membership functions and rule base are constructed;

[0009] Step b: Initialize the control model, including: setting the initial weights of the MPC model and the initial gains of the ESO; initializing the membership functions and rule base of the fuzzy adaptive module;

[0010] Step c: Fuse the data collected by the sensors to form the system state x(k), the actual measured output y(k), the system control input u(k), and the external disturbance intensity d(k);

[0011] Step d: Input x(k), u(k), and d(k) into the MPC model, predict the robot system state x(k+1) within the future sampling time domain N, and solve the MPC optimization problem that satisfies the system state x(k) constraint and the system control input u(k) constraint to obtain the optimal control sequence {u^(0),…,u^(N-1)} at time k;

[0012] Input x(k), u(k), d(k), y(k) into ESO, estimate the external disturbance in real time, and generate the estimated disturbance at time k

[0013] Step e: According to the optimal control sequence {u^(0),…,u^(N-1)}, obtain the initial system control input uMPC(k) required for the robot to maintain posture balance at time k; Used to compensate uMPC(k) and correct the system control input to the final control instruction ufinal(k); transmit the ufinal(k) to the motion actuator for robot posture control;

[0014] Step f: The robot posture error e(k), posture error change rate Δe(k) and regenerative disturbance intensity d m (k) Input the fuzzy adaptive module to update the MPC model weight and ESO gain; loop into the next cycle; the regenerative disturbance intensity d m (k) By estimating the disturbance generate.

[0015] Further preferably, the state space model is expressed as:

[0016] x(k+1)=Ax(k)+Bu(k)+Ed(k),

[0017] Among them, matrices A, B, and E are system matrices obtained through robot dynamics modeling;

[0018]

[0019] Among them, θ(k) is the robot attitude angle, including pitch angle, roll angle, and yaw angle; is the angular velocity of the robot's center of gravity; p(k) is the robot's position on the cable; v(k) is the robot's speed;

[0020] u(k)=[δ1(k),δ2(k),…,δ n (k),ω1(k),ω2(k),…,ω n (k)] T ,

[0021] Among them, δ i (k) represents the rotation angle of the robot's i-th mechanical leg, ω i (k) represents the speed of the robot's i-th running wheel; n is the number of the robot's mechanical legs and running wheels;

[0022]

[0023] Among them, d θ (k) is the change in robot posture angle caused by external disturbance, d p (k) is the change in robot position caused by external disturbances; the external disturbances include high-altitude wind and cable swing.

[0024] Still further preferably, the data collected by the sensors include: the robot's attitude angle, angular velocity, speed, acceleration and displacement collected in real time by the inertial measurement unit (IMU); the rotation angle of each mechanical leg collected in real time by the angle sensor; the walking wheel speed collected in real time by the speed sensor; and the external wind direction and wind speed collected in real time by the wind speed / direction sensor.

[0025] Still further preferably, the MPC optimization problem is expressed as:

[0026]

[0027] Where xref(k+i) is the expected state at time k+i, Q and R are the weight matrices of state error and control energy, respectively, and P is the terminal state weight.

[0028] Still further preferably, the input variables of the fuzzy adaptive module include:

[0029] e(k)=θ ref (k)-θ(k),

[0030] Δe(k)=e(k)-e(k-1),

[0031]

[0032] Among them, e(k) is the posture error, which represents the difference between the current actual posture of the system and the expected posture;

[0033] θ ref (k) is the expected posture; θ(k) is the actual measured posture; Δe(k) is the rate of change of posture error; e(k) is the posture error at time k; e(k-1) is the posture error at time k-1; d m (k) is the regenerative disturbance intensity at time k; is the estimated disturbance at time k;

[0034] The output variables of the fuzzy adaptive module include: an increment ΔQ or ΔR of the weight matrix Q or R of the MPC model; and a correction amount ΔL of the ESO gain matrix L.

[0035] Still further preferably, the extended state observer ESO is expressed as:

[0036]

[0037] in, is the extended state at time k; is the estimate of the extended state at time k; To expand the system matrix, its composition reflects the dynamics of the original system state x(k) and the assumed model of the disturbance d(k); is the input matrix of the extended system, which is used to describe the influence of the system control input on the extended state; u(k) represents the system control input applied at time k; L is the ESO gain matrix; y(k) is the actual measured output of the system at time k, and C is the mapping matrix from state to output; is the extended state estimate The part corresponding to state x(k).

[0038] Further preferably, the system state x(k) constraint includes the robot's spatial position constraint, motion speed constraint and posture angle constraint; the system control input u(k) constraint includes the robot's power actuator's speed constraint, acceleration constraint, torque constraint and displacement constraint.

[0039] Further preferably, a rolling horizon strategy is adopted to execute the optimal control sequence {u^(0),…,u^(N-1)}; only u^(0) is used to obtain the initial system control input uMPC(k) at the current moment.

[0040] Compared with the prior art, the present invention has the following beneficial effects:

[0041] 1. This is because the present invention effectively integrates MPC and ADRC: MPC is used to predict the future state of the robot's internal dynamics (such as leg lifting, speed change, and obstacle crossing), and optimize the corresponding control input signals to maintain the robot's posture balance; at the same time, the extended state observer (ESO) in ADRC is used to estimate external disturbances (such as high-altitude wind and wire swing) in real time to compensate for the control input signals, achieving dual compensation for internal factors and external disturbances, thereby ensuring the robot's posture balance under various complex working conditions.

[0042] 2. This invention introduces a fuzzy adaptive module based on fuzzy PID. This module uses real-time data collected from the attitude error, attitude error change rate, and disturbance intensity to optimize and update the MPC weight parameters and ESO gain in real time. This innovation significantly improves the system's response speed and robustness in the face of changing environments and sudden disturbances, overcoming the shortcomings of traditional fixed-parameter control methods.

[0043] 3. Because the present invention adopts a hierarchical control architecture, it takes into account the needs of global optimization and real-time disturbance compensation, and has good engineering feasibility. It not only improves the operating safety and control accuracy of the cable inspection robot, but also significantly reduces the overshoot and response delay problems, making the system more practical and adaptable in high-risk and high-complexity environments. Compared with the existing single control strategy, it has obvious performance advantages. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 This is a control flow chart of a cable inspection robot posture control method based on the fusion of MPC and ADRC in an embodiment of the present invention;

[0045] Figure 2 This is a schematic diagram of the control model structure of a cable inspection robot posture control method based on the fusion of MPC and ADRC in an embodiment of the present invention. DETAILED DESCRIPTION

[0046] In order to make the technical means, creative features, objectives and effects achieved by the present invention easy to understand, the present invention is specifically described in the following examples in conjunction with the accompanying drawings. It should be noted that the description of these implementation methods is used to help understand the present invention, but does not constitute a limitation of the present invention.

[0047] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art to which the present invention pertains. The terms used herein are for the purpose of describing specific embodiments only and are not intended to limit the present invention. It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this manner are interchangeable where appropriate so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions; the term "and / or" includes any and all combinations of one or more of the relevant listed items.

[0048] According to an embodiment of the present invention, an embodiment of a cable inspection robot posture control method based on the fusion of MPC and ADRC is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0049] The present invention proposes a posture control method for a cable inspection robot based on the fusion of MPC and ADRC. Aiming at the posture change problem of the cable inspection robot in a complex working environment due to two factors, internal actions (such as leg lifting, speed change, obstacle crossing) and external disturbances (such as high-altitude wind and cable swing), a hierarchical control scheme for the robot posture is proposed: first, a model predictive control (MPC) model is established to predict the future state of the robot and solve the optimal control problem to obtain the initial control input for realizing internal dynamic compensation; second, the extended state observer (ESO) in the active disturbance rejection control (ADRC) model is used to estimate the external disturbance in real time and use it to compensate the control input; third, a fuzzy adaptive module based on fuzzy PID is introduced to collect the posture error, posture error change rate and the regeneration disturbance intensity generated based on the estimated disturbance in real time, dynamically adjust the MPC weight and ESO gain, and improve the response speed and robustness of the system under variable working conditions. This technical solution effectively makes up for the deficiency of the existing single control method in maintaining the optimal control effect under variable working conditions, and provides a reliable technical guarantee for the stable operation of the cable inspection robot in high-risk and highly complex environments.

[0050] The present invention provides a cable inspection robot posture control method based on the fusion of MPC and ADRC, which includes the following contents and steps:

[0051] (1) Constructing the hardware system, including: building a hardware platform including a data processor, controller, communication equipment, power supply and power management module, and configuring sensor components.

[0052] (2) Establish a kinematic and dynamic model (state space model) about the relationship between the robot system state and internal dynamic input (system control input) and external disturbance.

[0053] (3) Based on the robot state space model established in step 2, the robot control system is modeled, including: internal dynamic control layer (MPC module), external disturbance compensation layer (ADRC module), and adaptive fusion layer (fuzzy adaptive module).

[0054] (4) Initialize the control model and input the data collected by each sensor in real time into the control system to obtain the control signal for the robot's posture balance. This mainly includes: after the sensor collects real-time status and disturbance data, the MPC module uses dynamic weights to predict the system state and output the optimal control sequence and corresponding system control input signal to maintain the robot's posture balance; the ADRC module estimates the disturbance in real time through ESO and generates the system control input compensation signal; the fuzzy adaptive module adjusts the MPC and ADRC parameters in real time based on the posture error, posture error change rate, and regenerative disturbance intensity. Finally, the output signals of the MPC module and ADRC module are fused to form a control instruction to drive the robot to adjust its posture. The entire process is executed cyclically within each sampling period to ensure stable system operation.

[0055] The following will introduce in detail the implementation details, algorithm derivation and system integration plan of each control system module.

[0056] 1. Hardware system construction

[0057] (1) Build the hardware platform, including:

[0058] Embedded data processor: Select a DSP or FPGA platform with high-speed computing capabilities to meet the requirements of MPC online solving and ESO real-time computing.

[0059] Communication interface: Configure CAN bus or Ethernet bus to ensure real-time data transmission between sensors, actuators and central controller.

[0060] Power supply and power management module: ensure stable power supply to each control unit and actuator.

[0061] Data fusion module: used to fuse (such as Kalman filtering) the data collected by each sensor and transmit the fused data to the controller.

[0062] (2) Configure sensor components to provide status feedback for the control algorithm, including:

[0063] Inertial Measurement Unit (IMU): Installed at the robot's geometric center of gravity (such as the chassis), it is used to collect the robot's attitude angle (pitch, roll, yaw), angular velocity, speed, acceleration, and displacement in real time.

[0064] Angle sensor: installed on the rotation joints of each robot leg (including walking legs and tower-climbing legs) to detect the rotation angle of each leg in real time.

[0065] Speed sensor: installed on the robot's walking mechanism (such as the walking wheel) to detect the walking wheel speed in real time.

[0066] Wind speed / direction sensor: Installed on the robot's external geometric center of gravity (such as the chassis), it is used to detect external wind direction and speed in real time, providing reference data for external disturbance estimation.

[0067] In this embodiment, the inertial measurement unit (IMU) is selected as the H30MINI model; the angle sensor is selected as the WT9011DCL-BT50 model; the speed sensor is selected as the IWT603 model; and the wind speed / direction sensor is selected as the LFM108 model.

[0068] 2. Kinematic and dynamic models of the robot

[0069] (1) Dynamic modeling and state space description

[0070] In order to realize closed-loop control, a dynamic model for robot posture control is first established.

[0071] The robot system state is defined as:

[0072]

[0073] Among them, θ(k) is the robot attitude angle (pitch angle, roll angle, and yaw angle can be modeled separately); is the angular velocity of the robot’s center of gravity; p(k) is the robot’s position on the cable (including lateral displacement); v(k) is the robot’s speed.

[0074] The robot system control input vector is defined as:

[0075] u(k)=[δ1(k),δ2(k),…,δ n (k),ω1(k),ω2(k),…,ω n (k)] T ,

[0076] Among them, δ i (k) represents the rotation angle of the robot's i-th mechanical leg, ω i(k) represents the speed of the robot's i-th running wheel; n is the number of the robot's mechanical legs and running wheels;

[0077] The disturbed state of the robot system is described as:

[0078]

[0079] Among them, d θ (k) is the change in robot posture angle caused by external disturbance, d p (k) is the change in robot position caused by external disturbances; external disturbances include high-altitude wind and cable swing.

[0080] In this embodiment, d θ (k), d p (k) are obtained through simulation experiments respectively. The input of the high-altitude wind disturbance simulation experiment is wind direction and wind speed, and the corresponding relationship between wind direction, wind speed and robot attitude angle and displacement disturbance is obtained; the input of the cable swing disturbance simulation experiment is the robot motion state (i.e. attitude angle, angular velocity, velocity, acceleration and displacement), and the robot attitude angle and displacement disturbance caused by the cable swing caused by the robot motion is obtained, that is, the corresponding relationship between the robot attitude angle, angular velocity, velocity, acceleration and displacement and the robot attitude angle and displacement disturbance is obtained. Therefore, the wind speed / wind direction sensor is input into the control system to obtain wind speed and wind direction, and the inertial measurement unit (IMU) obtains the robot attitude angle, angular velocity, velocity, acceleration and displacement, and thus d can be obtained. θ (k), d p (k).

[0081] Based on the Newton-Euler method, the spatial model of the robot system can be written as:

[0082] x(k+1)=Ax(k)+Bu(k)+Ed(k)

[0083] Matrix A describes the relationship between the system state variables, matrix B represents the effect of control inputs on the system state, and matrix E represents the effect of external disturbances on the system state. Matrices A, B, and E are derived based on the Newton-Euler method, taking into account the coupling effects of various actuators. Parameter identification can be performed using the Newton-Euler method and experimental data, and a discrete model can be obtained through discretization.

[0084] 3. Control system modeling

[0085] Based on the established robot dynamic model, the robot control system is modeled. Figure 1As shown in the figure, the entire control system is divided into three levels: (1) Internal dynamic control layer (MPC module): Using the dynamic model of the robot, the model predictive control (MPC) is used to optimize the internal input (system control input) (including the rotation angle of the robot's mechanical legs and the speed of the walking wheels) in real time, so that the robot can foresee and compensate for the posture deviation caused by internal factors when performing actions such as lifting legs and crossing obstacles. (2) External disturbance compensation layer (ADRC module): Design an extended state observer (ESO) to estimate and compensate for external disturbances such as high-altitude wind and cable swing in real time, so that the robot can quickly restore balance when affected by external factors. (3) Online adaptive fusion layer (fuzzy adaptive module): Using the idea of fuzzy PID, according to the real-time collected robot posture error, error change rate and disturbance intensity information, the MPC cost function weight parameters and ADRC observer gain are adjusted to ensure that the control system always maintains the best control performance under different working conditions. For MPC, the increment ΔQ or ΔR of the weight matrix Q or R is adjusted to control the posture changes caused by internal factors of the robot; for ADRC, the correction amount ΔL of the ESO gain is adjusted to control the posture changes caused by external factors of the machine.

[0086] (1) Internal dynamic control module (MPC design)

[0087] By constructing a prediction model and optimization problem, the robot system state x(k+1) within the future sampling time domain N is predicted, and the optimization problem is solved under the system state constraints and system control input constraints to obtain the optimal control sequence.

[0088] (a) Prediction model and optimization problem construction

[0089] Using the prediction time domain N (e.g. N = 10 sampling periods), construct the MPC optimization problem:

[0090]

[0091] Among them, xref(k+i) is the expected state at time k+i, matrices Q and R are the weight matrices of state error and control energy respectively, and P is the terminal state weight.

[0092] (b) Constraints

[0093] The control system must also satisfy the physical constraints of the control inputs and system states:

[0094] umin≤u(k+i)≤umax,i=0,…,N-1,

[0095] xmin≤x(k+i)≤xmax,i=0,…,N.

[0096] u(k+i) refers to the current control signal of the robot's power actuator (such as a motor or servo), that is, the system control input. Its constraints depend on the characteristics of the robot's power actuator, including: voltage constraints, speed constraints, acceleration constraints, torque constraints, position constraints, temperature constraints, etc.

[0097] x(k+i) represents the current state of the robot, and its constraints depend on the characteristics of the robot's motion mechanism, including the robot's spatial position constraints, motion speed constraints, and posture angle constraints.

[0098] (c) Solution and Execution

[0099] The QP solver is used to solve the above optimization problem in each sampling period to obtain the optimal control sequence {u^(0),…,u^(N-1)};

[0100] The optimal control sequence is executed using a rolling horizon strategy, which only uses u^(0) as the control instruction at the current moment, and then updates the state and solves the problem again in the next cycle.

[0101] The MPC module can predict the posture changes caused by internal actions (such as leg lifting and acceleration) in advance, and achieve smooth adjustment of the robot's posture.

[0102] (2) External disturbance compensation module (ADRC design)

[0103] like Figure 2 As shown in the figure, the ADRC module consists of three submodules: Tracking Differentiator (TD), Extended State Observer (ESO) and Nonlinear State Error Feedback (NLSEF). The Tracking Differentiator TD takes the robot's target posture (ideal posture) as input, uses a nonlinear differential algorithm to remove high-frequency noise in the reference signal, and outputs a smooth tracking posture signal and posture change rate, respectively, to ensure a smooth transition of subsequent control when the reference changes suddenly. The Extended State Observer ESO estimates the system's true state x(k) and the total disturbance it receives in real time online. Model uncertainty and external disturbances such as wind disturbances are aggregated into an extended state, and the estimation error is rapidly attenuated by designing appropriate observer gains. The nonlinear state error feedback (NLSEF) receives the outputs of the TD and ESO, and uses the robot's observed posture, observed posture error, and disturbance estimate as input. It uses a combination of nonlinear functions to generate the final compensation system control input, enabling the system to maintain stable tracking performance in the face of complex nonlinearities and strong disturbances. The controlled object is the robot itself, whose actual posture is fed into the ESO via sensors. The ESO's observations are then fed back to the NLSEF to form a closed-loop compensation loop and provide the TD with a reference filtered target trajectory. Together, these three ensure the high robustness and fast response of the control system.

[0104] (a) Extended State Observer (ESO) Design

[0105] Considering the system disturbance as an extended state, the extended state is defined as:

[0106]

[0107] The discretized form of ESO is:

[0108]

[0109] in,

[0110] is the extended state at time k;

[0111] is the estimate of the extended state at time k;

[0112] To expand the system matrix, its composition reflects the dynamics of the original system state x(k) and the assumed model of the disturbance d(k);

[0113] is the input matrix of the extended system, which is used to describe the influence of the control input on the extended state;

[0114] u(k) represents the control input applied at time k, such as the control signal of the rotation angle of the robot's mechanical leg motor and the speed of the walking wheel;

[0115] L is the ESO gain matrix, and its design goal is to ensure that the observer converges quickly enough and the estimation accuracy meets the requirements. In this embodiment, the appropriate convergence speed is determined by the pole placement method.

[0116] y(k) is the actual measured output of the system at time k, where C is the mapping matrix from state to output;

[0117] is the extended state estimate The part of that corresponds to the state x(k) (i.e., the first half of the vector).

[0118] (b) Disturbance compensation strategy

[0119] The disturbance estimated by ESO To compensate the MPC output, the corrected control input is:

[0120]

[0121] where uMPC(k) is the initial control input obtained by the MPC module at time k based on the optimal control sequence {u^(0),…,u^(N-1)} obtained by the system dynamics; is the disturbance estimated by the ESO at time k; K is the disturbance compensation gain matrix, which determines the compensation role of the disturbance estimate in the final control input; ufinal(k) represents the final control input after disturbance compensation at time k, that is, the posture control command, which will be sent to the power actuator to realize the robot posture control.

[0122] (3) Fuzzy adaptive module

[0123] A fuzzy control PID controller is designed. It inputs information about the posture error, error rate of change, and disturbance intensity. It outputs increments ΔQ or ΔR for the MPC cost function weight matrices Q and R, as well as a correction value ΔL for the ADRC observer gain. The MPC and ADRC modules are then updated to ensure that the control system maintains optimal control performance under various operating conditions. For MPC, the increments ΔQ or ΔR of the weight matrices Q or R are adjusted to control posture changes caused by internal factors. For ADRC, the ESO gain correction value ΔL is adjusted to control posture changes caused by external factors.

[0124] (a) Input and output design

[0125] Input variables:

[0126] Attitude error e(k) = θ ref (k)-θ(k),

[0127] Error change rate Δe(k) = e(k) - e(k-1),

[0128] Regeneration disturbance intensity

[0129] Where e(k) is the attitude error, which represents the difference between the current actual attitude of the system and the expected attitude; θ ref (k) is the expected posture; θ(k) is the actual measured posture.

[0130] Δe(k) is the error change rate, which describes the rate of change of the posture error between two consecutive time points and reflects the dynamic change trend of the error; e(k) is the posture error at time k; e(k-1) is the posture error at time k-1.

[0131] Regeneration disturbance intensity d m (k) By Obtained through vector operations.

[0132] Output variables:

[0133] For MPC, adjust the weight matrix Q or R by the increment ΔQ or ΔR;

[0134] For ADRC, the correction amount ΔL of the ESO gain is adjusted.

[0135] (b) Membership Function and Fuzzy Rules

[0136] According to the system characteristics, membership functions are designed for input variables, for example:

[0137] The error e is divided into "negative large", "negative small", "zero", "positive small", and "positive large";

[0138] The error change rate Δe is divided into "rapid rise", "slow rise", "stable", "slow fall", and "fast fall";

[0139] The disturbance intensity d is divided into "weak", "medium" and "strong".

[0140] Construct fuzzy rules, such as:

[0141] If e is positive and Δe is rising rapidly, the attitude deviation weight in Q is increased and the control penalty in R is reduced, prompting the MPC output to correct the deviation more actively;

[0142] If d is "strong", the ESO gain L is increased to speed up the disturbance estimation;

[0143] When both e and Δe are small and d is at a “weak” level, maintain or appropriately reduce the gain to avoid over-adjustment.

[0144] (c) Fuzzy Reasoning and Defuzzification

[0145] Use Mamdani or Sugeno type fuzzy inference to fuzzify each input, output fuzzy results according to the preset rule base, and then defuzzify through the center of gravity method to obtain accurate parameter correction values. Update formula example:

[0146] Q new =Q0+ΔQ,L new =L0+ΔL,

[0147] Among them, Q0 and L0 are initial parameters, and ΔQ and ΔL are real-time correction values calculated by the fuzzy module.

[0148] 4. System integration and control process

[0149] (1) Control process

[0150] (a) Initialization

[0151] Establish the robot state space model, set the MPC initial weights Q0, R0 and P, and the ESO initial gain L0; initialize the membership function and rule base of the fuzzy logic module.

[0152] (b) Real-time data collection

[0153] The data related to the robot's internal dynamics and external disturbances collected by each sensor are transmitted to the data fusion module. After data fusion, the system state x(k), actual measurement output y(k), system control input u(k), and external disturbance intensity d(k) are formed.

[0154] The data collected by the sensors include:

[0155] The inertial measurement unit (IMU) collects the robot's attitude angle (including pitch, roll, yaw), angular velocity, speed, acceleration and displacement in real time; the angle sensor collects the rotation angle of each mechanical leg in real time; the speed sensor collects the walking wheel speed in real time; the wind speed / direction sensor collects the external wind direction and wind speed in real time.

[0156] (c) MPC solution

[0157] According to the latest state and updated weight parameters, the MPC module is used to solve the optimal control sequence for the next N moments and output the corresponding uMPC(k).

[0158] (d) Perturbation estimation

[0159] In the ADRC module, the ESO updates the extended state estimate based on the current state and measurement data. Estimating external disturbances And output.

[0160] (e) Fuzzy adaptive adjustment

[0161] Calculate the current posture error e(k), error change rate Δe(k), and disturbance intensity d(k); the fuzzy logic module calculates the parameter adjustment amount based on the above inputs and updates the MPC Q, R and ESO gain L.

[0162] (f) Control signal fusion

[0163] The MPC output and the disturbance compensation signal are combined according to the formula Fusion to form the final control instructions.

[0164] (g) Implementation and Feedback

[0165] The final control instructions drive the speed of each walking wheel and the rotation angle of each leg respectively; during the execution process, the system response is monitored in real time and fed back to the controller to enter the next cycle.

[0166] (2) Experimental verification and parameter debugging

[0167] To verify the effectiveness of the control system, the following steps are performed on the experimental platform:

[0168] (a) Simulation test

[0169] Use MATLAB / Simulink or other simulation tools to build a cable inspection robot model and simulate the robot's posture changes under different working conditions (including internal motion changes and external disturbances);

[0170] The response speed, steady-state error and robustness of the fixed parameter and fuzzy adaptive control schemes are compared, and the membership function, fuzzy rules and control weights are adjusted.

[0171] (b) Hardware experiment

[0172] On the actual robot platform, a closed-loop control system is built to implement the above control process;

[0173] Experimental data feedback is used to further optimize MPC weights, ESO gains, and fuzzy adjustment parameters to ensure that the system achieves the expected performance in a practical environment.

[0174] (c) Data recording and analysis

[0175] The robot's posture response when crossing obstacles and being disturbed by wind is recorded, and the control effects of each module when used alone and in combination are compared to verify the comprehensive advantages of this patented solution.

[0176] This embodiment realizes the posture control of the cable inspection robot through the following steps: establishing an accurate multi-degree-of-freedom dynamic model to uniformly represent internal actions and external disturbances; using the MPC module to globally predict and optimize internal dynamics to ensure smooth adjustment of the robot in actions such as obstacle crossing and leg lifting; using the ESO designed in the ADRC module to estimate external disturbances in real time and compensate for control inputs; using the fuzzy adaptive adjustment module to adjust the MPC weights and ESO gains online according to the real-time operating status to ensure that the system has high response speed and robustness under different working conditions; verifying the effectiveness of the system design through simulation and actual testing, and continuously optimizing the parameters. This specific implementation method ensures that the cable inspection robot can still maintain stable posture control in complex dynamic environments, providing a comprehensive, accurate and logically clear technical solution for practical applications.

[0177] The above-mentioned embodiments are preferred cases of the present invention and are not intended to limit the scope of protection of the present invention. Various deformations or modifications that can be made by ordinary technicians in this field without creative work within the scope of the attached claims are still within the scope of protection of this patent.

Claims

1. A cable inspection robot posture control method based on the fusion of MPC and ADRC, characterized in that: The following steps are involved: Step a: Build a control model, including: Based on the Newton-Euler method, a state-space model is established for the robot system state x(k), the system control input u(k), and the external disturbance intensity d(k); based on the state-space model, an MPC model and an MPC optimization problem are constructed; based on the state-space model, an ADRC model and its extended state observer (ESO) are constructed; based on the fuzzy PID, a fuzzy adaptive module and its membership functions and rule base are constructed; Step b: Initialize the control model, including: setting the initial weights of the MPC model and the initial gains of the ESO; initializing the membership functions and rule base of the fuzzy adaptive module; Step c: Fuse the data collected by the sensors to form the system state x(k), the actual measured output y(k), the system control input u(k), and the external disturbance intensity d(k); Step d: Input x(k), u(k), and d(k) into the MPC model, predict the robot system state x(k+1) within the future sampling time domain N, and solve the MPC optimization problem that satisfies the system state x(k) constraint and the system control input u(k) constraint to obtain the optimal control sequence at time k. {u^(0),…,u^(N-1)}; Input x(k), u(k), d(k), y(k) into ESO, estimate the external disturbance in real time, and generate the estimated disturbance at time k Step e: According to the optimal control sequence {u^(0),…,u^(N-1)}, obtain the initial system control input uMPC(k) required for the robot to maintain posture balance at time k; Used to compensate uMPC(k) and correct the system control input to the final control instruction ufinal(k); transmit the ufinal(k) to the motion actuator for robot posture control; Step f: The robot posture error e(k), posture error change rate Δe(k) and regenerative disturbance intensity d m (k) Input the fuzzy adaptive module to update the MPC model weight and ESO gain; loop into the next cycle; the regenerative disturbance intensity d m (k) By estimating the disturbance generate.

2. The cable inspection robot posture control method based on the fusion of MPC and ADRC according to claim 1 is characterized in that: The state space model is expressed as: x(k+1)=Ax(k)+Bu(k)+Ed(k), Among them, matrices A, B, and E are system matrices obtained through robot dynamics modeling; Among them, θ(k) is the robot attitude angle, including pitch angle, roll angle, and yaw angle; is the angular velocity of the robot's center of gravity; p(k) is the robot's position on the cable; v(k) is the robot's speed; u(k)=[δ1(k),δ2(k),…,δ n (k),ω1(k),ω2(k),…,ω n (k)] T , Among them, δ i (k) represents the rotation angle of the robot's i-th mechanical leg, ω i (k) represents the speed of the robot's i-th running wheel; n is the number of the robot's mechanical legs and running wheels; Among them, d θ (k) is the change in robot posture angle caused by external disturbance, d p (k) is the change in robot position caused by external disturbances; the external disturbances include high-altitude wind and cable swing.

3. The cable inspection robot posture control method based on the fusion of MPC and ADRC according to claim 2 is characterized in that: The data collected by the sensors include: the robot's attitude angle, angular velocity, speed, acceleration and displacement collected in real time by the inertial measurement unit (IMU); the rotation angle of each mechanical leg collected in real time by the angle sensor; the walking wheel speed collected in real time by the speed sensor; and the external wind direction and wind speed collected in real time by the wind speed / direction sensor.

4. The cable inspection robot posture control method based on the fusion of MPC and ADRC according to claim 2 is characterized in that: The MPC optimization problem is expressed as: Where xref(k+i) is the expected state at time k+i, Q and R are the weight matrices of state error and control energy, respectively, and P is the terminal state weight.

5. The cable inspection robot posture control method based on the fusion of MPC and ADRC according to claim 2 is characterized in that: The extended state observer ESO is expressed as: in, is the extended state at time k; is the estimate of the extended state at time k; To expand the system matrix, its composition reflects the dynamics of the original system state x(k) and the assumed model of the disturbance d(k); is the input matrix of the extended system, which is used to describe the influence of the system control input on the extended state; u(k) represents the system control input applied at time k; L is the ESO gain matrix; y(k) is the actual measured output of the system at time k, and C is the mapping matrix from state to output; is the extended state estimate The part corresponding to state x(k).

6. The cable inspection robot posture control method based on the fusion of MPC and ADRC according to claim 4 is characterized in that: The input variables of the fuzzy adaptive module include: e(k)=θ ref (k)-θ(k), Δe(k)=e(k)-e(k-1), Among them, e(k) is the posture error, which represents the difference between the current actual posture of the system and the expected posture; θ ref (k) is the desired posture; θ(k) is the actual measured posture; Δe(k) is the rate of change of posture error; e(k) is the posture error at time k; e(k-1) is the posture error at time k-1; d m (k) is the regenerative disturbance intensity at time k; is the estimated disturbance at time k; The output variables of the fuzzy adaptive module include: The increment ΔQ or ΔR of the weight matrix Q or R of the MPC model; the correction amount ΔL of the ESO gain matrix L.

7. The cable inspection robot posture control method based on the fusion of MPC and ADRC according to claim 1 is characterized in that: The system state x(k) constraints include robot spatial position constraints, motion speed constraints and posture angle constraints; The system control input u(k) constraints include speed constraints, acceleration constraints, torque constraints, and displacement constraints of the robot power actuator.

8. The cable inspection robot posture control method based on the fusion of MPC and ADRC according to claim 1 is characterized in that: The optimal control sequence {u^(0),…,u^(N-1)} is executed using the rolling horizon strategy; only u ^ (0) is used to obtain the initial system control input uMPC(k) at the current moment.

Images