Underwater robot dynamic positioning method for coping with complex disturbance

Abstract

The invention provides an underwater robot dynamic positioning method for coping with complex disturbance, and the method comprises the steps: collecting the attitude angle, angular velocity, linear velocity and linear acceleration information of an underwater robot, and generating fused high-frequency and high-precision pose information; designing an extended state disturbance observer, and estimating a disturbance value in real time through a feedback function and an auxiliary variable equation; adding the disturbance estimation value and the output force of the dynamic model, constructing a constraint condition of model prediction control, and solving the optimal control input of N prediction time domains in the future through rolling time domain optimization; and based on the optimal control input, constructing a quadratic programming problem of thrust distribution in combination with the space pose distribution matrix of the thrusters, solving the minimum energy solution of the thrusters, and outputting the minimum energy solution to a thruster execution mechanism. According to the technical scheme of the invention, the robust control of the underwater robot in a strong disturbance environment can be realized. The method mainly solves the problem of fixed point and attitude determination when the underwater unmanned robot faces underwater turbulence or other complex disturbances.

Description

Technical Field

[0001] The present invention belongs to the technical field of autonomous control of unmanned underwater robots, and in particular relates to a dynamic positioning method for underwater robots that cope with complex disturbances. Background Art

[0002] In recent years, the operational tasks and scenarios of underwater robots have become increasingly complex, leading to several technical challenges, particularly in motion control. First, the underwater environment is highly unpredictable, with disturbances such as undercurrents and ocean currents. Second, accurate path tracking in a highly damped environment presents challenges.

[0003] The dynamic coupling forces of underwater robots are complex, and the parameters of their dynamic models have significant uncertainty. Dynamic models are currently the most widely used model-based control. There are two main approaches to building dynamic models: knowledge-driven modeling and data-driven modeling. Knowledge-driven modeling involves constructing a spatial motion model based on Newton's laws of motion and mathematical physics. A commonly used model is Fossen's AUV and ROV fluid dynamics model. This model can be further simplified based on the actual operating environment and payload configuration of the underwater robot.

[0004] Based on established models, the most widely used control methods internationally are model-based control methods such as model predictive control and sliding mode control. For difficult-to-measure state variables such as modeling errors and external disturbances, model-based disturbance observers are often used to observe and estimate these control variables. While underwater positioning of underwater robots can be achieved to a certain extent, problems exist such as inaccuracy, instability, and difficulty in implementation. Summary of the Invention

[0005] In response to the problems existing in the prior art, the present invention provides a dynamic positioning method for underwater robots that can cope with complex disturbances, which can achieve robust control of underwater robots in strong disturbance environments and solve the problem of fixed point and attitude determination of underwater unmanned robots when facing underwater turbulence or other complex disturbances.

[0006] In an embodiment of the present invention, a method for dynamic positioning of an underwater robot for coping with complex disturbances is provided, comprising:

[0007] S101, collecting the attitude angle, angular velocity, linear velocity, and linear acceleration information of the underwater robot through an inertial measurement unit (IMU) and a binocular camera, calibrating the gyroscope noise and accelerometer bias of the IMU and the posture measurement noise of the binocular camera using an error state Kalman filter algorithm, and generating fused high-frequency and high-precision posture information;

[0008] S102. Based on the six-degree-of-freedom dynamic model of the underwater robot, an extended state disturbance observer (ESO) is designed. External unknown disturbances and internal parameter changes are modeled as extended state variables, and the disturbance value is estimated in real time through feedback functions and auxiliary variable equations.

[0009] S103, summing the disturbance estimate and the output force of the dynamic model, constructing the constraints of the model predictive control (MPC), defining the loss function as the weighted sum of squares of the state tracking error and the quadratic term of the control input, and solving the optimal control input for the next N prediction horizons through rolling horizon optimization;

[0010] S104. Based on the optimal control input, combined with the spatial posture distribution matrix of the thrusters, and with the maximum thrust of each thruster as an inequality constraint, a quadratic programming problem for thrust distribution is constructed, and the minimum energy solution of the thrust of each thruster is solved using the Lagrange multiplier method, and the solution is output to the thruster actuator.

[0011] Furthermore, the implementation of the error state Kalman filter algorithm includes:

[0012] Calibrate the IMU's gyroscope angle random walk coefficient, accelerometer bias instability parameter, and stereo camera's image feature point matching noise variance through experiments, and write the calibration parameters into a YAML-formatted configuration file.

[0013] Construct a state space equation with attitude quaternion error, velocity error, and position error as state quantities. Its state transfer matrix includes the angular velocity integral term of the IMU and the gravity compensation term of the accelerometer.

[0014] The Kalman gain matrix is ​​dynamically adjusted according to the sampling frequency of the binocular camera so that the frequency of the fused pose data is not lower than the original sampling frequency of the IMU.

[0015] Furthermore, the design of the extended state disturbance observer satisfies the following conditions:

[0016] Assume that the external disturbance force has weak time-varying characteristics within the sampling period, that is, the first-order derivative of the disturbance force with time is approximately zero;

[0017] Auxiliary variables are defined as linear combinations of observer states, whose dynamic equations satisfy in, is an auxiliary variable, K1 and K2 are diagonal positive definite feedback matrices, x is the actual state of the system, Estimate the state for the observer.

[0018] Furthermore, the loss function of the MPC controller is defined as:

[0019]

[0020] Among them, x(k) is the state vector of the kth step in the prediction time domain, including position, velocity and attitude angle; x ref (k) is the expected state at the corresponding moment, Q is the diagonal weight matrix, and its diagonal elements correspond to the penalty coefficients of position error, velocity error, and attitude error; R is the diagonal weight matrix, which controls the quadratic penalty term coefficient of the input increment; Δu(k) controls the input increment to avoid violent fluctuations.

[0021] Furthermore, the method further includes a dynamic disturbance compensation mechanism:

[0022] When the binocular camera loses its pose data due to underwater turbidity, it switches to the short-term pose prediction mode of the pure inertial measurement unit (IMU) and compensates for the positioning error caused by the missing visual data through the extended state disturbance observer (ESO) observer.

[0023] Furthermore, based on the assumption of statistical weak time-varying properties of high-intensity flow fields, the time derivative of the disturbance force is approximately regarded as 0, and an extended state disturbance observer is designed.

[0024] The specific design method of ESO disturbance observer includes: taking feedback function P(v) = L(v)v; setting auxiliary variable z to satisfy Disturbance observer estimate It can be expressed as

[0025] The tracking error e satisfies According to the assumption that the disturbance force is weakly time-varying, the derivative of the disturbance force is considered to be 0. Design L as a constant matrix, and we can get It can be seen that it eventually converges to 0.

[0026] Furthermore, the system control force and the external disturbance force are added together to obtain the actual resultant force that the thruster needs to output. This resultant force is then integrated into the constraints and loss function of the MPC to obtain a more physically meaningful optimal solution. The specific form is:

[0027] Introducing the weight matrices Q and W, and the target state R, we define the following loss function:

[0028]

[0029] in, is the extended position matrix; the constraints can be expressed as: Among them, λ is the scaling factor, F is the limit thrust in each direction; by minimizing the loss function, the control quantity of the next N prediction time domains is obtained Each operation only takes the control quantity at time k as the control input, thereby realizing the rolling optimization process.

[0030] Furthermore, the method also includes an energy consumption optimization strategy, namely:

[0031] Update the weight coefficient of the Lagrange multiplier method online based on historical thrust distribution data, giving priority to reducing the frequency of use of high-power thrusters;

[0032] When it is detected that the battery power is lower than the threshold, the position tracking accuracy constraints are automatically relaxed to reduce the total energy consumption and extend the battery life.

[0033] Furthermore, the fusion positioning process of the inertial measurement unit (IMU) includes:

[0034] The ORB-SLAM algorithm is used to construct a real-time map of underwater environment feature points, and the camera coordinate system is converted to the IMU body coordinate system through the hand-eye calibration matrix;

[0035] Using the prediction-update loop of the error state Kalman filter, visual pose corrections are inserted between the IMU high-frequency data to generate a fused pose output above 200Hz.

[0036] The beneficial effects brought about by the present invention are as follows:

[0037] As can be seen from the above scheme, the embodiment of the present invention provides a dynamic positioning method for underwater robots to cope with complex disturbances. By collecting the attitude angle, angular velocity, linear velocity and linear acceleration information of the underwater robot, the gyroscope noise, accelerometer bias of the IMU and the posture measurement noise of the binocular camera are calibrated to generate fused high-frequency and high-precision posture information; the external unknown disturbance and internal parameter changes are modeled as extended state variables, and the disturbance value is estimated in real time through the feedback function and the auxiliary variable equation; the disturbance estimate is summed with the output force of the dynamic model to construct the constraint conditions of the model predictive control, and the loss function is defined by the weighted square sum of the state tracking error and the quadratic term of the control input. The optimal control input for the next N predicted time domains is solved by rolling time domain optimization; based on the optimal control input, combined with the spatial posture distribution matrix of the thruster, the maximum thrust of each thruster is used as the inequality constraint, and the quadratic programming problem of thrust distribution is constructed to solve the minimum energy solution of the thrust of each thruster and output it to the thruster actuator. The technical solution of the present invention can realize the robust control of underwater robots in a strong disturbance environment. The present invention mainly solves the problem of fixed point and fixed attitude of underwater unmanned robots when facing underwater turbulence or other complex disturbances. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 This is a flow chart of a dynamic positioning method for an underwater robot to cope with complex disturbances according to an embodiment of the present invention;

[0039] Figure 2This is a schematic diagram of the effect of an extended state disturbance observer of a dynamic positioning method for an underwater robot coping with complex disturbances according to an embodiment of the present invention;

[0040] Figure 3 Schematic diagram comparing the dynamic positioning effect of an underwater robot dynamic positioning method for coping with complex disturbances according to an embodiment of the present invention and a traditional control method. DETAILED DESCRIPTION

[0041] To make the objectives, technical solutions, and advantages of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts shall fall within the scope of protection of the present invention.

[0042] like Figures 1 to 3 As shown, Figure 1 This is a flow chart of a method for dynamic positioning of an underwater robot to cope with complex disturbances according to an embodiment of the present invention. Figure 2 Schematic diagram of the effect of an extended state disturbance observer of an underwater robot dynamic positioning method for coping with complex disturbances according to an embodiment of the present invention. Figure 3 Schematic diagram comparing the dynamic positioning effect of an underwater robot dynamic positioning method for coping with complex disturbances according to an embodiment of the present invention and a traditional control method.

[0043] Figure 1 A dynamic positioning method for an underwater robot to cope with complex disturbances includes:

[0044] S101, collecting the attitude angle, angular velocity, linear velocity, and linear acceleration information of the underwater robot through an inertial measurement unit (IMU) and a binocular camera, calibrating the gyroscope noise and accelerometer bias of the IMU and the posture measurement noise of the binocular camera using an error state Kalman filter algorithm, and generating fused high-frequency and high-precision posture information;

[0045] S102. Based on the six-degree-of-freedom dynamic model of the underwater robot, an extended state disturbance observer (ESO) is designed. External unknown disturbances and internal parameter changes are modeled as extended state variables, and the disturbance value is estimated in real time through feedback functions and auxiliary variable equations.

[0046] S103, summing the disturbance estimate and the output force of the dynamic model, constructing the constraints of the model predictive control (MPC), defining the loss function as the weighted sum of squares of the state tracking error and the quadratic term of the control input, and solving the optimal control input for the next N prediction horizons through rolling horizon optimization;

[0047] S104. Based on the optimal control input, combined with the spatial posture distribution matrix of the thrusters, and with the maximum thrust of each thruster as an inequality constraint, a quadratic programming problem for thrust distribution is constructed, and the minimum energy solution of the thrust of each thruster is solved using the Lagrange multiplier method, and the solution is output to the thruster actuator.

[0048] In an embodiment of the present invention, a dynamic positioning method for an unmanned underwater robot that copes with complex disturbances uses IMU and binocular camera information as its data foundation. Sensor data fusion is achieved through internal parameter calibration, improving the accuracy and return frequency of the underwater robot's short-term internal position and velocity data. The assumption of weak time-varying statistics of high-intensity flow fields is introduced in the process of constructing a disturbance observer, improving the mathematical rigor of the convergence proof process. An MPC controller is used to calculate the control system input. The disturbance observer observation results are integrated in this process to obtain a more physically meaningful optimal control force input. The Lagrange multiplier method is used to distribute thrust based on the spatial distribution of the thrusters and the goal of minimizing energy consumption, thereby increasing the robot's flight time.

[0049] In one embodiment of the present invention, the implementation of the error state Kalman filter algorithm includes:

[0050] Calibrate the IMU's gyroscope angle random walk coefficient, accelerometer bias instability parameter, and stereo camera's image feature point matching noise variance through experiments, and write the calibration parameters into a YAML-formatted configuration file.

[0051] Construct a state space equation with attitude quaternion error, velocity error, and position error as state quantities. Its state transfer matrix includes the angular velocity integral term of the IMU and the gravity compensation term of the accelerometer.

[0052] The Kalman gain matrix is ​​dynamically adjusted according to the sampling frequency of the binocular camera so that the frequency of the fused pose data is not lower than the original sampling frequency of the IMU.

[0053] In one embodiment of the present invention, the design of the extended state disturbance observer satisfies the following conditions:

[0054] Assume that the external disturbance force has weak time-varying characteristics within the sampling period, that is, the first-order derivative of the disturbance force with time is approximately zero;

[0055] Auxiliary variables are defined as linear combinations of observer states, whose dynamic equations satisfy in, is an auxiliary variable, K1 and K2 are diagonal positive definite feedback matrices, x is the actual state of the system, Estimate the state for the observer.

[0056] In one embodiment of the present invention, the loss function of the MPC controller is defined as:

[0057]

[0058] Among them, x(k) is the state vector of the kth step in the prediction time domain, including position, velocity and attitude angle; x ref (k) is the expected state at the corresponding moment, Q is the diagonal weight matrix, and its diagonal elements correspond to the penalty coefficients of position error, velocity error, and attitude error; R is the diagonal weight matrix, which controls the quadratic penalty term coefficient of the input increment; Δu(k) controls the input increment to avoid violent fluctuations.

[0059] In one embodiment of the present invention, the method further includes a dynamic disturbance compensation mechanism:

[0060] When the binocular camera loses its pose data due to underwater turbidity, it switches to the short-term pose prediction mode of the pure inertial measurement unit (IMU) and compensates for the positioning error caused by the missing visual data through the extended state disturbance observer (ESO) observer.

[0061] In one embodiment of the present invention, based on the assumption of statistical weak time-varying properties of high-intensity flow fields, the time derivative of the disturbance force is approximately regarded as 0, and an extended state disturbance observer is designed;

[0062] The specific design method of ESO disturbance observer includes: taking feedback function P(v) = L(v)v; setting auxiliary variable z to satisfy Disturbance observer estimate It can be expressed as

[0063] The tracking error e satisfies According to the assumption that the disturbance force is weakly time-varying, the derivative of the disturbance force is considered to be 0. Design L as a constant matrix, and we can get It can be seen that it eventually converges to 0.

[0064] In one embodiment of the present invention, the system control force and the external disturbance force are added to obtain the actual resultant force that the thruster needs to output. This resultant force is then integrated into the constraints and loss function of the MPC to obtain a more physically meaningful optimal solution. The specific form is:

[0065] Introducing the weight matrices Q and W, and the target state R, we define the following loss function:

[0066]

[0067] in, is the extended position matrix; the constraints can be expressed as: Among them, λ is the scaling factor, F is the limit thrust in each direction; by minimizing the loss function, the control quantity of the next N prediction time domains is obtained Each operation only takes the control quantity at time k as the control input, thereby realizing the rolling optimization process.

[0068] In one embodiment of the present invention, the method further includes an energy consumption optimization strategy, namely:

[0069] Update the weight coefficient of the Lagrange multiplier method online based on historical thrust distribution data, giving priority to reducing the frequency of use of high-power thrusters;

[0070] When it is detected that the battery power is lower than the threshold, the position tracking accuracy constraints are automatically relaxed to reduce the total energy consumption and extend the battery life.

[0071] In one embodiment of the present invention, the fusion positioning process of the inertial measurement unit (IMU) includes:

[0072] The ORB-SLAM algorithm is used to construct a real-time map of underwater environment feature points, and the camera coordinate system is converted to the IMU body coordinate system through the hand-eye calibration matrix;

[0073] Using the prediction-update loop of the error state Kalman filter, visual pose corrections are inserted between the IMU high-frequency data to generate a fused pose output above 200Hz.

[0074] In this embodiment of the present invention, the designed MPC controller integrates disturbance observations. Unlike the method of directly subtracting the disturbance observations from the MPC output control force, the observations are utilized by summing the system control force with the external disturbance force to obtain the actual thruster output force. This force is then integrated into the MPC constraints and loss function to obtain a more physically meaningful optimal solution.

[0075] In one embodiment of the present invention, a dynamic positioning method for an unmanned underwater robot in response to complex disturbances is provided. The dynamic positioning method for an underwater robot in response to complex disturbances is described with reference to the accompanying drawings.

[0076] First, calibrate the noise and random walk of the IMU and binocular camera, and write the parameters to a YAML file. Use the error state Kalman filter algorithm to fuse sensor information and construct a state-space equation based on acceleration, velocity, and attitude information. During data sampling, update the Kalman gain in real time based on the sampling frequency to improve sensor perception accuracy and data return rate.

[0077] Based on the statistical weak time-varying assumption of high-intensity flow field, the time derivative of disturbance force is approximately regarded as 0, and the extended state disturbance observer is designed. The specific design method of ESO disturbance observer is: take the feedback function P(v) = L(v)v; set the auxiliary variable z to satisfy Disturbance observer estimate It can be expressed as The proof is as follows: The tracking error e satisfies According to the assumption that the disturbance force is weakly time-varying, the derivative of the disturbance force is considered to be 0. Design L as a constant matrix, and we can get It can be seen that it eventually converges to 0. Figure 2 As shown, Figure 2 A schematic diagram of the effect of the extended state disturbance observer used in the present invention in the simulation environment is given.

[0078] Based on the disturbance force observation results, an MPC controller is designed. Unlike the method of directly subtracting the disturbance observation value from the MPC output control force, the control system observation results are used by adding the system control force and the external disturbance force to obtain the actual thruster output force. This force is then integrated into the MPC constraints and loss function to obtain a more physically meaningful optimal solution. The specific form is: introduce the weight matrices Q and W, and the target state R, and define the following loss function in is the extended position matrix; the constraints can be expressed as: Where λ is the scaling factor and F is the limit thrust in each direction. By minimizing the loss function, the control quantity for the next N prediction time domains can be obtained. Each operation only takes the control quantity at time k as the control input, thereby realizing the rolling optimization process.

[0079] Based on the control input calculation results, the Lagrange multiplier method is used, with the ultimate thrust of the thruster as the constraint condition and the sum of the squares of the thrusts of each thruster as the loss function. The thrust of each thruster is calculated and the result is applied to the control system to complete the dynamic positioning control in a disturbance environment. Figure 3 As shown, Figure 3 Schematic diagram comparing the dynamic positioning effect of the present invention and the traditional control method.

[0080] An embodiment of the present invention provides a dynamic positioning method for an underwater robot that copes with complex disturbances. The method collects the attitude angle, angular velocity, linear velocity and linear acceleration information of the underwater robot, calibrates the gyroscope noise, accelerometer bias of the IMU and the posture measurement noise of the binocular camera, and generates fused high-frequency and high-precision posture information. External unknown disturbances and internal parameter changes are modeled as extended state variables, and the disturbance value is estimated in real time through feedback functions and auxiliary variable equations. The disturbance estimate is summed with the output force of the dynamic model to construct the constraint conditions of the model predictive control, and the loss function is defined by the weighted sum of squares of the state tracking error and the quadratic term of the control input. The optimal control input for the next N predicted time domains is solved through rolling time domain optimization. Based on the optimal control input, combined with the spatial posture distribution matrix of the thruster, the maximum thrust of each thruster is used as an inequality constraint to construct a quadratic programming problem for thrust distribution, solve the minimum energy solution of the thrust of each thruster, and output it to the thruster actuator.

[0081] The technical solution of the present invention can achieve robust control of underwater robots in highly disturbed environments. The present invention mainly solves the problem of fixed point and fixed attitude of underwater unmanned robots when facing underwater turbulence or other complex disturbances.

[0082] The above is a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A dynamic positioning method for underwater robots to cope with complex disturbances, characterized in that: The method comprises: S101, collecting the attitude angle, angular velocity, linear velocity, and linear acceleration information of the underwater robot through an inertial measurement unit (IMU) and a binocular camera, calibrating the gyroscope noise and accelerometer bias of the IMU and the posture measurement noise of the binocular camera using an error state Kalman filter algorithm, and generating fused high-frequency and high-precision posture information; S102. Based on the six-degree-of-freedom dynamic model of the underwater robot, an extended state disturbance observer (ESO) is designed. External unknown disturbances and internal parameter changes are modeled as extended state variables, and the disturbance value is estimated in real time through feedback functions and auxiliary variable equations. S103, summing the disturbance estimate and the output force of the dynamic model, constructing the constraints of the model predictive control (MPC), defining the loss function as the weighted sum of squares of the state tracking error and the quadratic term of the control input, and solving the optimal control input for the next N prediction horizons through rolling horizon optimization; S104. Based on the optimal control input, combined with the spatial posture distribution matrix of the thrusters, and with the maximum thrust of each thruster as an inequality constraint, a quadratic programming problem for thrust distribution is constructed, and the minimum energy solution of the thrust of each thruster is solved using the Lagrange multiplier method, and the solution is output to the thruster actuator.

2. The method for dynamic positioning of an underwater robot for coping with complex disturbances according to claim 1, characterized in that: The implementation of the error state Kalman filter algorithm includes: Calibrate the IMU's gyroscope angle random walk coefficient, accelerometer bias instability parameter, and stereo camera's image feature point matching noise variance through experiments, and write the calibration parameters into a YAML-formatted configuration file. Construct a state space equation with attitude quaternion error, velocity error, and position error as state quantities. Its state transfer matrix includes the angular velocity integral term of the IMU and the gravity compensation term of the accelerometer. The Kalman gain matrix is ​​dynamically adjusted according to the sampling frequency of the binocular camera so that the frequency of the fused pose data is not lower than the original sampling frequency of the IMU.

3. The method for dynamic positioning of an underwater robot for coping with complex disturbances according to claim 1, characterized in that: The design of the extended state disturbance observer satisfies the following conditions: Assume that the external disturbance force has weak time-varying characteristics within the sampling period, that is, the first-order derivative of the disturbance force with time is approximately zero; Auxiliary variables are defined as linear combinations of observer states, whose dynamic equations satisfy in, is an auxiliary variable, K1 and K2 are diagonal positive definite feedback matrices, x is the actual state of the system, Estimate the state for the observer.

4. The method for dynamic positioning of an underwater robot for coping with complex disturbances according to claim 1, characterized in that: The loss function of the MPC controller is defined as: Among them, x(k) is the state vector of the kth step in the prediction time domain, including position, velocity and attitude angle; x ref (k) is the expected state at the corresponding moment, Q is the diagonal weight matrix, and its diagonal elements correspond to the penalty coefficients of position error, velocity error, and attitude error; R is the diagonal weight matrix, which controls the quadratic penalty term coefficient of the input increment; Δu(k) controls the input increment to avoid violent fluctuations.

5. The method for dynamic positioning of an underwater robot for coping with complex disturbances according to claim 1, characterized in that: The method further includes a dynamic disturbance compensation mechanism: When the binocular camera loses its pose data due to underwater turbidity, it switches to the short-term pose prediction mode of the pure inertial measurement unit (IMU) and compensates for the positioning error caused by the missing visual data through the extended state disturbance observer (ESO) observer.

6. The method for dynamic positioning of an underwater robot for coping with complex disturbances according to claim 1, characterized in that: Based on the statistical weak time-varying assumption of high-intensity flow field, the time derivative of the disturbance force is approximately regarded as 0, and an extended state disturbance observer is designed. The specific design method of ESO disturbance observer includes: taking feedback function P(v) = L(v)v; setting auxiliary variable z to satisfy Disturbance observer estimate It can be expressed as The tracking error e satisfies According to the assumption that the disturbance force is weakly time-varying, the derivative of the disturbance force is considered to be 0. Design L as a constant matrix, and we can get It can be seen that it eventually converges to 0.

7. The method for dynamic positioning of an underwater robot for coping with complex disturbances according to claim 1, characterized in that: The system control force and the external disturbance force are added together to obtain the actual force output by the thruster. This force is then integrated into the constraints and loss function of the MPC to obtain a more physically meaningful optimal solution. The specific form is: Introducing the weight matrices Q and W, and the target state R, we define the following loss function: in, is the extended position matrix; the constraints can be expressed as: Among them, λ is the scaling factor, F is the limit thrust in each direction; by minimizing the loss function, the control quantity of the next N prediction time domains is obtained Each operation only takes the control quantity at time k as the control input, thereby realizing the rolling optimization process.

8. The method for dynamic positioning of an underwater robot for coping with complex disturbances according to claim 1, characterized in that: The method further includes an energy consumption optimization strategy, namely: Update the weight coefficient of the Lagrange multiplier method online based on historical thrust distribution data, giving priority to reducing the frequency of use of high-power thrusters; When it is detected that the battery power is lower than the threshold, the position tracking accuracy constraints are automatically relaxed to reduce the total energy consumption and extend the battery life.

9. The method for dynamic positioning of an underwater robot for coping with complex disturbances according to claim 1, characterized in that: The fusion positioning process of the inertial measurement unit (IMU) includes: The ORB-SLAM algorithm is used to construct a real-time map of underwater environment feature points, and the camera coordinate system is converted to the IMU body coordinate system through the hand-eye calibration matrix; Using the prediction-update loop of the error state Kalman filter, visual pose corrections are inserted between the IMU high-frequency data to generate a fused pose output above 200Hz.

Images