A method and system for adaptive disturbance compensation for deep-sea robotic propulsion control

Abstract

This invention provides an adaptive disturbance compensation propulsion control method and system for deep-sea robots, relating to the field of deep-sea robotics. The method includes: collecting deep-sea environmental data around the robot; constructing a rigid-body-hydrodynamic model of the robot carrying environmental disturbances based on the deep-sea environmental data to describe the robot's multi-degree-of-freedom motion in the deep-sea environment; acquiring commands for each thruster of the robot; introducing a bilinear fault model of multiplicative failure and additive jamming into the thruster commands; calculating the actual output thrust of each thruster; calculating the robot's comprehensive disturbance based on the bilinear fault model and environmental disturbances; performing sparse Gaussian process learning on the robot's comprehensive disturbance based on actual robot detection data; outputting the mean and variance of the comprehensive disturbance; performing stochastic model predictive control based on the mean and variance of the comprehensive disturbance; calculating the optimal virtual control resultant force of the stochastic model predictive control using a cost-minimizing method; and executing the action commands corresponding to the optimal virtual control resultant force.

Description

Technical Field

[0001] This invention relates to the field of deep-sea robot technology, and in particular to an adaptive disturbance compensation propulsion control method and system for deep-sea robots. Background Technology

[0002] Currently, traditional deep-sea robot control methods mainly rely on PID control and independent fault detection and isolation techniques. Traditional PID control suffers from the following technical problems: 1. Traditional PID control schemes cannot handle strong nonlinearity and coupled control, and have weak anti-interference capabilities. In the high-pressure environment of the 6000-meter deep sea, this can easily lead to micro-deformation of the robot structure and changes in buoyancy. The unknown ocean currents, eddies, and geological changes in the deep-sea environment also generate low-frequency large disturbances that significantly impact robot control. Furthermore, the slender robotic arms or payloads of deep-sea robots experience severe perturbations in model parameters, making it difficult for traditional PID control schemes to establish accurate dynamic models. Additionally, the high pressure, low temperature, and sediment erosion in the deep sea often cause propeller entanglement, bearing wear, and seal failure, resulting in jamming, efficiency loss, or characteristic drift, and these gradual faults are difficult to detect in their early stages. 2. In traditional technical solutions, independent fault detection and isolation techniques rely on accurate residual thresholds, leading to high false alarm and false negative rates under large disturbances in the deep sea. Moreover, the control is intermittent during reconfiguration, making the final control effect prone to deviating from the expected target. Summary of the Invention

[0003] One of the objectives of this invention is to provide an adaptive disturbance compensation propulsion control method and system for deep-sea robots. The method and system construct a rigid body-hydrodynamic model of the robot that incorporates deep-sea environmental disturbances, including water flow disturbances and geological disturbances in the deep-sea environment. Therefore, the rigid body-hydrodynamic model modeled by this invention can effectively describe the motion model in the complex deep-sea environment. This allows the invention to consider the impact of different disturbances on propulsion control in the actual propulsion control process, effectively improving the anti-interference effect of the deep-sea robot on the control target.

[0004] Another objective of this invention is to provide an adaptive disturbance compensation propulsion control method and system for deep-sea robots. The method and system construct a dual-line fault model based on multiplicative failure and additive jamming for each thruster of the robot in a deep-sea environment. Furthermore, this invention calculates the comprehensive disturbance of the deep-sea robot based on the dual-line fault model of multiplicative failure and additive jamming, combined with environmental disturbances. Therefore, in the actual control process, this invention considers environmental disturbances and robot fault disturbances as superimposed factors to ultimately affect the propulsion results, thereby significantly improving the prediction of the impact of various uncertainties on actual propulsion control in a deep-sea environment.

[0005] Another objective of this invention is to provide an adaptive perturbation compensation method and system for propulsion control of deep-sea robots. This method and system utilizes a sparse Gaussian process (GP) for online perturbation learning based on a real-world deep-sea robot detection dataset. After learning using the sparse Gaussian process, the system outputs a Gaussian-distributed mean and variance of the comprehensive perturbation. This invention can then use the mean and variance of the comprehensive perturbation to provide time-varying and probabilistic feedforward compensation for a stochastic model predictive controller, thereby generating an optimal virtual resultant force that satisfies constraints. Therefore, this invention can address both the propulsion control accuracy and the control safety issues caused by uncertain perturbations in deep-sea environments.

[0006] Another objective of this invention is to provide an adaptive disturbance compensation propulsion control method and system for deep-sea robots. After acquiring the state of the deep-sea robot, the method and system construct opportunity constraints based on the safety boundary of the state, and use the stochastic model predictive controller to convert the opportunity constraints into deterministic contraction constraints. This enables the deep-sea robot of this invention to automatically contract the safety boundary under uncertain conditions with large variances, effectively isolating high-risk actions.

[0007] To achieve at least one of the above-mentioned objectives, the present invention further provides an adaptive disturbance compensation propulsion control method for a deep-sea robot, the method comprising:

[0008] Collect deep-sea environmental data around the robot, and construct a rigid body-hydrodynamic model of the robot carrying environmental disturbances based on the deep-sea environmental data to describe the robot's multi-degree-of-freedom motion in the deep-sea environment;

[0009] Obtain each thruster command of the robot, and introduce a dual-line fault model of multiplicative failure and additive jamming into the thruster command. Calculate the actual output thrust of each thruster based on the dual-line fault model and the thruster command.

[0010] The robot's overall disturbance is calculated based on the dual-line fault model and environmental disturbances. Based on the robot's actual detection data, the robot's overall disturbance is learned through a sparse Gaussian process, and the mean and variance of the overall disturbance are output.

[0011] Stochastic model predictive control is performed based on the mean and variance of the comprehensive disturbance, and the optimal virtual control resultant force of the stochastic model predictive control is calculated by minimizing the cost, and the action command corresponding to the optimal virtual control resultant force is executed.

[0012] According to a preferred embodiment of the present invention, the method for constructing the rigid body-hydrodynamic model of the robot carrying environmental disturbances includes: constructing a volume coordinate system with the center point of the robot body as the origin, and constructing an inertial coordinate system based on the volume coordinate system; obtaining the position and attitude vectors of the robot in the corresponding coordinate system, obtaining the velocity vector in the corresponding volume coordinate system, and calculating the acceleration; and calculating the system inertial force matrix, Coriolis centripetal force matrix, hydrodynamic damping matrix, and restoring force matrix based on the position, attitude vector, velocity vector, and acceleration of the robot, which are used to superimpose and describe the six-degree-of-freedom control force matrix containing environmental disturbances.

[0013] According to another preferred embodiment of the present invention, the dual-path fault model construction method includes: obtaining the commanded thrust of each thruster; obtaining the actual thrust of executing the commanded thrust based on the force sensor of each thruster; dividing the actual thrust by the commanded thrust as the multiplicative effectiveness coefficient of the thruster; and further monitoring the propeller state of each thruster; if there is a change in the commanded thrust of the corresponding thruster, but the current propeller state of the corresponding thruster remains unchanged, generating a jamming bias of the corresponding thruster based on the actual thrust value of the original propeller state; multiplying the multiplicative effectiveness coefficient and the commanded thrust as a multiplicative failure term; using the jamming bias as an additive jamming term; and combining the multiplicative failure term and the additive jamming term to describe the dual-path fault model to obtain the actual thrust of the corresponding thruster in the fault state.

[0014] According to another preferred embodiment of the present invention, after calculating the actual thrust of each thruster based on the multiplicative failure term and the additive jamming term, the actual thrust of each thruster is used as a vector to obtain the actual resultant thrust of the robot; further, a comprehensive disturbance including environmental disturbance and fault model disturbance is calculated, and the multidimensional comprehensive disturbance is added to the robot rigid-hydrodynamic model, so that the robot thrust control under fault conditions can be completed based on the robot rigid-hydrodynamic model considering only one unknown concentrated quantity.

[0015] According to another preferred embodiment of the present invention, the sparse Gaussian process learning method for the comprehensive disturbance of the robot includes: given an input thrust command, modeling the output as a Gaussian distribution of the comprehensive disturbance using the sparse Gaussian process, further calculating the similarity between two input points using a squared exponential kernel as a kernel function, and obtaining the mean and variance of the comprehensive disturbance of all thrusters through the sparse Gaussian process learning prediction, inputting the mean of the comprehensive disturbance of all thrusters into the corresponding robot rigid body-hydrodynamic model for overall compensation, and performing stochastic model predictive control based on the variance and mean to output the action command of the optimal virtual control resultant force.

[0016] According to another preferred embodiment of the present invention, the stochastic model predictive control method includes: establishing a discretized hydrodynamic model with actual velocity as the target control variable based on the sampling time, and establishing chance constraints containing multiple safety boundaries, further converting the chance constraints into deterministic contraction constraints, wherein the deterministic contraction constraints automatically adjust the boundary constraint range according to the size of the disturbance variance, and automatically shrink the constraint range of the corresponding boundary when the disturbance variance increases, in order to isolate high-risk areas or actions and improve the safety of the predicted command thrust action.

[0017] According to another preferred embodiment of the present invention, the cost function calculation method in the stochastic model predictive control method includes: acquiring the state vectors of the robot's real-time position, attitude, and velocity, and calculating the robot's state tracking error based on the reference position, reference attitude, and reference velocity; calculating the penalty term for incremental control to suppress drastic changes in control commands; and calculating the robot's terminal cost, wherein the terminal cost is used to lock the state within the executable region; constructing the cost function of the stochastic model predictive control based on the state tracking error, the penalty term for incremental control, and the terminal cost; and minimizing the cost function to obtain the action command of the optimal virtual control resultant force.

[0018] According to another preferred embodiment of the present invention, a penalty usage rule for each thruster is configured based on the multiplicative effectiveness coefficient of the thruster. When the multiplicative effectiveness coefficient of the thruster is lower, the usage penalty of the corresponding thruster is increased. The remaining thrust demand obtained based on the usage penalty is allocated to other thrusters to generate thrust commands with corresponding thrust.

[0019] To achieve at least one of the above-mentioned objectives, the present invention further provides an adaptive disturbance compensation propulsion control system for a deep-sea robot, wherein the system executes the aforementioned adaptive disturbance compensation propulsion control method for a deep-sea robot.

[0020] The present invention further provides a computer-readable storage medium storing a computer program, which is executed by a processor to implement the above-described adaptive disturbance compensation propulsion control method for deep-sea robots. Attached Figure Description

[0021] Figure 1 The diagram shown is a flowchart of an adaptive disturbance compensation propulsion control method for deep-sea robots according to the present invention. Detailed Implementation

[0022] The following description is intended to disclose the present invention and enable those skilled in the art to implement it. The preferred embodiments described below are merely examples, and other obvious variations will occur to those skilled in the art. The basic principles of the invention defined in the following description can be applied to other embodiments, modifications, improvements, equivalents, and other technical solutions that do not depart from the spirit and scope of the invention.

[0023] It is understood that the term "a" should be understood as "at least one" or "one or more," that is, in one embodiment, the number of an element can be one, while in another embodiment, the number of the element can be multiple, and the term "a" should not be understood as a limitation on the number.

[0024] Please combine Figure 1 This invention discloses an adaptive disturbance compensation propulsion control method and system for deep-sea robots, wherein the method mainly includes the following steps:

[0025] S01. Collect deep-sea environmental data around the robot, and construct a robot rigid body-hydrodynamic model carrying environmental disturbances based on the deep-sea environmental data to describe the robot's multi-degree-of-freedom motion in the deep-sea environment.

[0026] S02. Obtain the command for each thruster of the robot, and introduce a dual-line fault model of multiplicative failure and additive jamming into the thruster command. Calculate the actual output thrust of each thruster based on the dual-line fault model and the thruster command.

[0027] S03. Calculate the robot's comprehensive disturbance based on the dual-line fault model and environmental disturbances. Based on the robot's actual detection data, perform sparse Gaussian process learning on the robot's comprehensive disturbance and output the mean and variance of the comprehensive disturbance.

[0028] S04. Perform stochastic model predictive control based on the comprehensive disturbance mean and variance, calculate the optimal virtual control resultant force of the stochastic model predictive control by minimizing the cost, and execute the action command corresponding to the optimal virtual control resultant force.

[0029] Specifically, the deep-sea environmental data described in this invention can utilize sensors including, but not limited to, acoustic Doppler flow sensors, vibration sensors, and image sensors to detect water flow data and geological change data in the deep-sea environment. The water flow data includes stable and periodic constant currents and tidal currents, as well as unstable eddies and turbulent currents. The vibration data can collect geological change data, including seismic waves transmitted from the seabed. The image sensors can collect and identify geological change data related to seafloor volcanic eruptions and lava flows. This geological change data can be used to construct a safety boundary. One of the core technical aspects of this invention is that it uses the robot's environmental data in the deep-sea environment as perturbation data to construct a rigid-body-hydrodynamic model of the robot. Therefore, the rigid-body-hydrodynamic model of the robot carrying environmental perturbations in this invention can more accurately describe the influencing variables in real-world scenarios, thus providing multi-dimensional and accurate reference factors for subsequent precise control of the thrusters.

[0030] The method for constructing the rigid-body-hydrodynamic model of the robot carrying environmental disturbances includes: firstly, constructing a body coordinate system with the robot's center point as the origin, and then constructing a convertible inertial coordinate system based on the body coordinate system; acquiring the robot's position and pose data and the corresponding robot velocity in the constructed coordinate system; calculating the robot's acceleration at the velocity; and using the robot's position data, pose data, velocity data, and acceleration data as input variables to construct the rigid-body-hydrodynamic model of the robot, wherein the rigid-body-hydrodynamic model of the robot carrying environmental disturbances includes: an inertial matrix M, a Coriolis-centripetal force matrix C(v), a hydrodynamic damping matrix D(v), and a buoyancy-restoring force matrix. The inertia matrix M, the Coriolis-centripetal force matrix C(v), the hydrodynamic damping matrix D(v), and the buoyancy restoring force matrix are... Superimposed description of thruster resultant force and environmental disturbance Therefore, the formula for the rigid body-hydrodynamic model of the robot carrying environmental disturbances is:

[0031] ;

[0032] in Indicates position and attitude. These represent data for different coordinate axes in the inertial coordinate system. These represent the roll angle, pitch angle, and yaw angle, respectively. Where v on the left represents the velocity vector. These represent the longitudinal sway, transverse sway, and heave linear velocities in the body coordinate system, respectively. These represent the roll, pitch, and yaw angular velocities, respectively. This represents the derivative of velocity, i.e., acceleration. , Represents the moment of inertia of a rigid body. The inertia matrix M represents the fluid-added mass inertia and is used to represent the total inertial force that the robot needs to overcome during acceleration. The resultant force of the thrusters is described below. The calculation formulas include:

[0033] ;

[0034] in This represents the thruster configuration matrix, with dimensions 6×m, where m represents the number of thrusters and 6 represents the thruster degrees of freedom. The thruster configuration matrix maps the thrust commands of each thruster to a six-dimensional resultant force / torque on the aircraft. Matrix elements consist of the thruster installation position and thrust direction angle. Decision. In the formula. This represents the commanded thrust vector of each thruster.

[0035] Furthermore, this invention constructs a dual-line fault model based on multiplicative failure and additive jamming. It should be noted that due to the unique deep-sea environment, each thruster may be affected by factors including, but not limited to, environmental or internal mechanical wear, which may prevent each thruster from fully executing the target thrust action; that is, the final actual thrust may be less than the target thrust. Therefore, this invention designs the aforementioned multiplicative failure fault line, where the fault characteristics of the multiplicative failure may be slowly and gradually change, thus requiring repeated detection in subsequent tests. The multiplicative effectiveness coefficient related to the multiplicative failure can be obtained by comparing the actual thrust value detected by the pressure sensor connected and installed inside the thruster with the thrust command value.

[0036] In deep-sea environments, system malfunctions may lead to partial thruster jamming, where the actual propeller does not change its movement when the thruster command changes, and remains in its original state. It should be noted that the original state of the jammed thruster can be a zero-speed state or a specific thrust state. The state will be determined by monitoring the propeller state of each thruster according to the jamming type, which will be used to construct the jamming bias of the dual-line fault model.

[0037] Therefore, based on the multiplicative effectiveness coefficient and the jamming bias, the present invention obtains the following bilinear fault model:

[0038] ;

[0039] in This represents the actual thrust output of the i-th thruster. This represents the commanded thrust of the i-th thruster. This represents the multiplicative effective coefficient of the i-th thruster, with a value range of 1. ,in Indicates a state of perfect health. Indicates complete failure, 0 This indicates a loss of efficiency. This represents the additive jamming bias of the i-th thruster, indicating that the thruster is jammed at a certain thrust value, and the thrust cannot be adjusted even if the command changes. Further calculations are performed on the actual output thrust vector of all thrusters. :

[0040] ;

[0041] in This represents a diagonal matrix whose diagonal elements are the multiplicative effective coefficients of each thruster. , m represents the number of thrusters. Therefore, the actual resultant force generated... It can be represented as:

[0042] ;

[0043] The combined disturbance d is further calculated as follows:

[0044] ;

[0045] in This indicates the deviation in effective thrust caused by the loss of thruster efficiency. The effect of jamming bias on the resultant force. After simplification using the hydrodynamic model, the following robot rigid-body-hydrodynamic damping model, incorporating environmental and fault disturbances, is obtained: It should be noted that It includes a combination of environmental and fault disturbances.

[0046] Furthermore, the present invention is also based on the aforementioned integrated disturbance. Perform sparse Gaussian process learning, wherein the sparse Gaussian process learning will synthesize the perturbation The j-th component is modeled as a Gaussian process, wherein the Gaussian process uses velocity and inference instructions as input feature vectors, and the input feature vectors are defined as follows: ,in The input feature vector is used to predict the current integrated perturbation. The formula for the Gaussian modeling process is: ,in Representing a Gaussian process, given by the prior mean function Sum of squares kernel Confirmed, among which

[0047] in Representing the signal variance, it is a hyperparameter of the control function amplitude. This represents the length scaling matrix, which determines the rate of correlation decay for each input dimension. Represents the noise variance, describing the variance of observed noise. Represent the Kronecker function, when A value of 1 is used if the condition is met, and 0 otherwise. This is added to the diagonal of the kernel to account for independent noise. This results in the following combined disturbance mean and variance being output for each control cycle, where the combined disturbance mean... The calculation formula is: The above formula describes the test point The average of the comprehensive disturbance predictions at the location The combined disturbance variance The calculation formula is: The above formula describes the test point Comprehensive disturbance prediction variance .

[0048] Furthermore, in the stochastic model predictive control method of this invention, it is necessary to consider the sampling time. Discretizing the robot rigid-body-hydrodynamic damping model, which includes environmental and fault disturbances, with speed control as the objective, yields the following formula:

[0049] ;

[0050] in, Indicates discrete sampling time. and Let these represent the velocity vectors at step k and step (k+1), respectively. This represents the thrust command vector at step k. This represents the mean of the comprehensive disturbance prediction at step k. , and Let these represent the Coriolis force matrix, hydrodynamic damping matrix, and restoring force vector corresponding to the k-th step, respectively. B represents the inverse of the inertia matrix, and B represents the thruster matrix.

[0051] Further construct the following cost function :

[0052] ;

[0053] in, The expected cost is represented by , and the control objective in the stochastic model predictive control is to minimize the cost; N represents the prediction time domain length and the number of deviations. Represents the state tracking error, where It is a complete state vector containing position, attitude, and velocity. It is reference trajectory data for position, attitude, and velocity, used for comparison with actual data. This represents the weighted squared error. Represents the state error weight matrix. This indicates a control increment, used to suppress drastic changes in instructions. This represents a penalty term that controls the increment. This represents the control increment weight matrix. Let P represent the terminal cost, and let P represent the terminal weight matrix, satisfying the relevant Lyapunov equation.

[0054] It is worth mentioning that the chance constraint calculation formula with multiple boundaries constructed based on the stochastic model predictive control in this invention is as follows:

[0055] in Represents probability. This represents the robot state at step k. This represents the set of safe state boundaries, which includes, but is not limited to, position boundaries, velocity limit boundaries, etc. Indicates the confidence level. This represents the upper limit of the tolerance probability of default. The formula for transforming the above opportunity constraint into a deterministic contraction constraint is:

[0056] ;

[0057] H and h represent the set of security state boundaries. Matrices and vectors represented as linear inequalities; This represents the predicted mean of the state at step k. The prediction covariance matrix representing the state at step k can be obtained by integrating perturbation variance propagation; This represents the inverse standard normal cumulative distribution function, used to determine the degree of constraint tightening. It represents the projection of the standard deviation of the state uncertainty caused by the disturbance onto the direction of the constraint normal, i.e., the contraction.

[0058] Furthermore, the present invention also constructs a penalty usage rule for each thruster based on the multiplicative effectiveness coefficient of the thruster. The lower the multiplicative effectiveness coefficient of the thruster, the higher the usage penalty for the corresponding thruster. For example, when the multiplicative effectiveness coefficient of the thruster is lower than a certain threshold, the usage penalty will reset the current thruster usage to zero, and the remaining thrust demand obtained based on the usage penalty will be allocated to other thrusters to generate corresponding thrust propulsion commands.

[0059] The processes described in the flowcharts above, as disclosed in the embodiments of this invention, can be implemented as computer software programs. The embodiments disclosed in this invention include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication component, and / or installed from a removable medium. When the computer program is executed by a central processing unit (CPU), the methods of this application are not limited to the aforementioned functions. It should be noted that the computer-readable medium described above in this application can be a computer-readable signal medium or a computer-readable storage medium, or any combination of the two. Computer-readable storage media can be, for example, but not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatuses, or devices, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to: electrical connections having one or more conductor segments, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this application, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in connection with an instruction execution system, apparatus, or device. In this application, a computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A computer-readable signal medium can also be any computer-readable medium other than a computer-readable storage medium, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on a computer-readable medium can be transmitted using any suitable medium, including but not limited to: wireless, wired, optical fiber, RF, etc. Or any suitable combination of the above.

[0060] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation that may be implemented in systems, methods, and computer program products according to various embodiments of the present invention. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, may be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0061] Those skilled in the art should understand that the embodiments of the present invention described above and shown in the accompanying drawings are merely examples and do not limit the present invention. The purpose of the present invention has been fully and effectively achieved. The functions and structural principles of the present invention have been shown and explained in the embodiments. Without departing from the stated principles, the implementation of the present invention may have any variations or modifications.

Claims

1. A propulsion control method for a deep-sea robot with adaptive disturbance compensation, characterized in that, The method includes: Collect deep-sea environmental data around the robot, and construct a rigid body-hydrodynamic model of the robot carrying environmental disturbances based on the deep-sea environmental data to describe the robot's multi-degree-of-freedom motion in the deep-sea environment; Obtain each thruster command of the robot, and introduce a dual-line fault model of multiplicative failure and additive jamming into the thruster command. Calculate the actual output thrust of each thruster based on the dual-line fault model and the thruster command. The robot's overall disturbance is calculated based on the dual-line fault model and environmental disturbances. Based on the robot's actual detection data, the robot's overall disturbance is learned through a sparse Gaussian process, and the mean and variance of the overall disturbance are output. Stochastic model predictive control is performed based on the mean and variance of the comprehensive disturbance, and the optimal virtual control resultant force of the stochastic model predictive control is calculated by minimizing the cost, and the action command corresponding to the optimal virtual control resultant force is executed.

2. The adaptive disturbance compensation propulsion control method for deep-sea robots according to claim 1, characterized in that, The method for constructing the rigid body-hydrodynamic model of the robot carrying environmental disturbances includes: constructing a volume coordinate system with the center point of the robot body as the origin, and constructing an inertial coordinate system based on the volume coordinate system; obtaining the position and attitude vectors of the robot in the corresponding coordinate system, obtaining the velocity vector in the corresponding volume coordinate system, and calculating the acceleration; and calculating the system inertial force matrix, Coriolis centripetal force matrix, hydrodynamic damping matrix, and restoring force matrix based on the position, attitude vector, velocity vector, and acceleration of the robot, which are used to superimpose and describe the six-degree-of-freedom control force matrix containing environmental disturbances.

3. The adaptive disturbance compensation propulsion control method for deep-sea robots according to claim 1, characterized in that, The dual-path fault model construction method includes: acquiring the commanded thrust of each thruster; acquiring the actual thrust of executing the commanded thrust based on the force sensor of each thruster; dividing the actual thrust by the commanded thrust as the multiplicative effectiveness coefficient of the thruster; and further monitoring the propeller state of each thruster. If there is a change in the commanded thrust of the corresponding thruster, but the current propeller state of the corresponding thruster remains unchanged, a jamming bias of the corresponding thruster is generated using the actual thrust value of the original propeller state. The multiplicative effectiveness coefficient and the commanded thrust are multiplied together to obtain a multiplicative failure term, and the jamming bias is obtained as an additive jamming term. The dual-path fault model is described by combining the multiplicative failure term and the additive jamming term to obtain the actual thrust of the corresponding thruster under fault conditions.

4. The adaptive disturbance compensation propulsion control method for deep-sea robots according to claim 3, characterized in that, After calculating the actual thrust of each thruster based on the multiplicative failure term and the additive jamming term, the actual thrust of each thruster is used as a vector to obtain the actual resultant thrust of the robot. Further calculation is performed on the comprehensive disturbance, which includes environmental disturbance and fault model disturbance. The multidimensional comprehensive disturbance is added to the robot rigid-hydrodynamic model, so that the robot's thrust control under fault conditions can be completed by considering only one unknown concentrated quantity based on the robot rigid-hydrodynamic model.

5. The adaptive disturbance compensation propulsion control method for deep-sea robots according to claim 1, characterized in that, The sparse Gaussian process learning method for the robot's comprehensive disturbance includes: given an input thrust command, the output is modeled as a Gaussian distribution of the comprehensive disturbance using the sparse Gaussian process, and the similarity between two input points is calculated using a squared exponential kernel as the kernel function. After learning and predicting the mean and variance of the comprehensive disturbance of all thrusters through the sparse Gaussian process, the mean of the comprehensive disturbance of all thrusters is input into the corresponding robot rigid body-hydrodynamic model for overall compensation. Based on the variance and mean, stochastic model predictive control is performed to output the action command of the optimal virtual control resultant force.

6. The adaptive disturbance compensation propulsion control method for deep-sea robots according to claim 1, characterized in that, The stochastic model predictive control method includes: establishing a discretized hydrodynamic model with actual velocity as the target control variable based on the sampling time, and establishing chance constraints containing multiple safety boundaries. The chance constraints are further transformed into deterministic contraction constraints, wherein the deterministic contraction constraints automatically adjust the boundary constraint range according to the size of the disturbance variance. When the disturbance variance increases, the constraint range of the corresponding boundary is automatically reduced to isolate high-risk areas or actions and improve the safety of the predicted command thrust actions.

7. The adaptive disturbance compensation propulsion control method for deep-sea robots according to claim 1, characterized in that, The cost function calculation method in the stochastic model predictive control method includes: acquiring the robot's real-time position, attitude, and velocity state vectors, and calculating the robot's state tracking error based on reference position, reference attitude, and reference velocity; calculating the robot's penalty term for incremental control to suppress drastic changes in control commands; and calculating the robot's terminal cost, wherein the terminal cost is used to lock the state within the executable region; constructing the cost function of the stochastic model predictive control based on the state tracking error, the penalty term for incremental control, and the terminal cost; and minimizing the cost function to obtain the action command of the optimal virtual control resultant force.

8. The adaptive disturbance compensation propulsion control method for deep-sea robots according to claim 1, characterized in that, The penalty usage rules for each thruster are configured according to the multiplicative effectiveness coefficient of the thruster. The lower the multiplicative effectiveness coefficient of the thruster, the higher the usage penalty for the corresponding thruster. The remaining thrust demand obtained based on the usage penalty is allocated to other thrusters to generate thrust commands with corresponding thrust.

9. An adaptive disturbance compensation propulsion control system for a deep-sea robot, characterized in that, The system executes an adaptive disturbance compensation propulsion control method for deep-sea robots as described in any one of claims 1-8.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that is executed by a processor to implement the adaptive disturbance compensation propulsion control method for a deep-sea robot as described in any one of claims 1-8.

Images