UMV man-machine cooperation asynchronous H-infinity control method based on fuzzy Semi-Markov theoretical modeling

Abstract

The invention relates to the technical field of ship control and navigation, and discloses a UMV man-machine cooperation asynchronous H-infinity control method based on fuzzy Semi-Markov theoretical modeling, and the method comprises the following steps: constructing a T-S fuzzy Semi-Markov system model of an unmanned ship UMV; based on the T-S fuzzy Semi-Markov system model, the structure of a man-machine cooperation asynchronous controller is designed; the man-machine cooperation asynchronous controller is substituted into the T-S fuzzy Semi-Markov system model, and a closed-loop control system is obtained; solving the matrix inequality criterion by using a linear matrix inequality technology; and deploying the complete man-machine cooperation asynchronous controller to the unmanned ship. According to the method, through modeling and man-machine cooperation asynchronous control, accurate characterization of a non-exponential distribution working condition, robust stability under modal mismatch and excellent maneuvering performance fused with expert experience are realized.

Description

Technical Field

[0001] This invention relates to the field of ship control and navigation technology, specifically to a UMV human-machine collaborative asynchronous H∞ control method based on fuzzy Semi-Markov theory modeling. Background Technology

[0002] As the core carrier of future intelligent shipping, unmanned vessels are playing an increasingly important role in fields such as marine resource exploration, environmental monitoring, and military reconnaissance. Because ships face extremely complex marine environments during navigation, external disturbances such as wind, waves, and currents exhibit strong randomness and nonlinearity. Furthermore, the ship's own load conditions (such as full load or empty load) or draft change with the progress of the mission, causing unmanned vessels to exhibit significant nonlinear dynamic characteristics and multimodal switching features.

[0003] To describe this complex dynamic behavior, academia and industry often use a combination of TS fuzzy models and Markov Jump Systems (MJSs) for modeling. This method uses TS fuzzy logic to approximate the nonlinearity of the system and uses Markov processes to characterize the switching between different operating modes. However, traditional Markov jump system models rely on a strong theoretical assumption: the system's dwell time in a particular mode must follow an exponential distribution, exhibiting "memorylessness." But in actual maritime practice, the duration of sea states or the holding time of specific mission conditions often follows a Weibull distribution or other non-exponential distributions, exhibiting significant memory characteristics. Forcing the use of traditional Markov models to describe this behavior would lead to deviations between the model parameters and the actual physical processes, thus affecting the accuracy and reliability of the controller.

[0004] Furthermore, modern unmanned vessels typically operate based on networked control systems, requiring sensor-collected status data to be transmitted to the controller via wireless networks. Due to limited communication bandwidth, signal interference, or distance factors, network-induced latency or packet loss is unavoidable in data transmission. This results in the controller receiving modal information that often lags behind the vessel's current actual physical mode, leading to an asynchronous phenomenon where the "controller mode" and "system mode" are mismatched. Existing synchronous control strategies mostly assume that the controller can acquire system modes in real time and accurately. Once modal mismatch occurs, the stability of the closed-loop system becomes difficult to guarantee, and may even lead to loss of control of the vessel.

[0005] At the design level of control strategies, existing H∞ control or robust control methods mostly focus on pure mathematical numerical optimization, aiming to obtain the optimal control gain by solving linear matrix inequalities. Although this method can theoretically guarantee system stability, in engineering applications, the control law calculated by pure numerical solutions is prone to producing excessively large or high-frequency steering responses, which does not conform to actual navigation habits and can also exacerbate wear on actuators. The operational experience accumulated by senior captains or maritime experts in long-term practice (such as stabilization strategies under specific sea conditions) is of great value in improving the stability and safety of the system, but existing automated control frameworks often struggle to effectively integrate this unstructured prior knowledge into standardized controller design processes. Summary of the Invention

[0006] To address the shortcomings of existing technologies, this invention provides a UMV human-machine collaborative asynchronous H∞ control method based on fuzzy Semi-Markov theory modeling. This method solves the problems in existing unmanned ship control, such as the distortion in modeling non-exponential operating conditions caused by the limitations of traditional Markov assumptions, the asynchronous mismatch between system and controller modes caused by network communication delays, and the lack of expert experience guidance in control strategies that rely solely on numerical optimization, resulting in insufficient practicality of operation.

[0007] To achieve the above objectives, the present invention provides the following technical solution: The first aspect of this invention provides a UMV human-machine cooperative asynchronous H∞ control method based on fuzzy Semi-Markov theory modeling, comprising the following steps: S1. Construct a TS fuzzy Semi-Markov system model for an unmanned vessel (UMV). The system model uses fuzzy logic to approximate the nonlinear dynamic characteristics of the unmanned vessel and uses a semi-Markov process to describe the random switching characteristics between system modes. S2. Based on the TS fuzzy Semi-Markov system model, design the structure of the human-machine collaborative asynchronous controller, set the total control gain of the asynchronous controller to be superimposed by the machine calculation gain to be determined and the preset human experience gain, and define the asynchronous mapping relationship between the controller mode and the system mode. S3. Substitute the human-machine collaborative asynchronous controller into the TS fuzzy Semi-Markov system model to obtain a closed-loop control system. Construct a Lyapunov function that depends on the system mode and derive the matrix inequality criterion that makes the closed-loop control system satisfy the stochastic stability and H∞ performance index. S4. Solve the matrix inequality criterion using linear matrix inequality techniques to obtain a numerical solution for the machine computation gain, and combine the numerical solution with the human experience gain to generate a complete human-machine collaborative asynchronous controller. S5. Deploy the complete human-machine collaborative asynchronous controller to the unmanned vessel, calculate the control input based on the real-time collected status data, and perform navigation control on the unmanned vessel.

[0008] Preferably, the process of constructing a TS fuzzy Semi-Markov system model for an unmanned vessel (UMV) includes: Define the system state vector, control input signal vector, external disturbance vector, and performance output vector of the unmanned vessel. The system state vector includes the displacement and velocity information of the vessel in the pitch, sway, and yaw directions. Set the number of fuzzy rules and define corresponding prerequisite variables and membership functions for each fuzzy rule. For each system mode determined by a semi-Markov process, establish a local linear state-space equation corresponding to the fuzzy rule. The local linear state-space equation includes the system matrix, input matrix, and disturbance matrix related to the current system mode. Use the membership function to perform a weighted summation of all local linear state-space equations to obtain a global TS fuzzy Semi-Markov system model that can describe the nonlinear dynamic characteristics of the unmanned vessel under different operating conditions.

[0009] Preferably, the process of describing the stochastic switching characteristics between system modes using a semi-Markov process includes: The probability density function is used to describe the residence time distribution of the system modes in a certain state. The residence time distribution follows a non-exponential distribution, which makes the transition rate between modes a time-varying function that depends on the residence time. Calculate the expected value of the time-varying transfer rate function to obtain the constantized transfer rate matrix; By introducing the system mode index as a variable into the correlation matrix parameters of the system model, the dynamic parameters of the system model change abruptly with the random switching of system modes, thereby characterizing the dynamic characteristics of unmanned ships under different loads or marine environments.

[0010] Preferably, the process of designing the structure of the human-machine collaborative asynchronous controller includes: designing the fuzzy rule structure of the controller using a parallel distributed compensation strategy, so that it shares the same premise variables and membership functions with the TS fuzzy Semi-Markov system model; defining the local control law form of the controller, with each local control law corresponding to a fuzzy rule; The gain matrix structure of the local control law is set to an additive combination form, that is, it is formed by directly adding a machine-calculated gain matrix to be solved and a pre-set human experience gain matrix. The final control input expression is obtained by weighting and summing the local control law containing superimposed gains using membership functions.

[0011] Preferably, the process of defining the asynchronous mapping relationship between controller modes and system modes includes: The system modal index is set to represent the current actual physical mode of the unmanned vessel, and the controller modal index is set to represent the modal information received by the controller. An asynchronous description mechanism is constructed, which allows the controller modal index and the system modal index to have different values, thereby characterizing the modal mismatch between the controller and the system caused by sensor measurement delay, signal transmission delay or data packet loss; In controller design, both machine computational gain and human experience gain are made dependent on the controller modal index and fuzzy rule index, rather than directly dependent on the system modal index.

[0012] In one specific embodiment, the human experience gain is obtained as follows: Based on historical maneuvering data of unmanned vessels under specific sea conditions or expert knowledge bases, determine the fixed gain values ​​under different controller modes and fuzzy rules; The fixed gain values ​​are constructed as a known constant matrix and injected into the controller structure as prior knowledge to constrain or guide the direction of the machine's gain calculation during the optimization process. The machine-calculated gain is then used as a decision variable to be optimized, and its specific value is determined through the subsequent linear matrix inequality solution process.

[0013] Preferably, the process of deriving the matrix inequality criterion that enables the closed-loop control system to satisfy the stochastic stability and H∞ performance index includes: A stochastic Lyapunov function dependent on the system mode index is constructed. The stochastic Lyapunov function is defined as a quadratic form of the system state vector and a sequence of positive definite matrices to be determined, with the positive definite matrix sequence taking different values ​​as the system mode index changes. Weak infinitesimal operator operations are performed on the stochastic Lyapunov function to calculate the differential expectation of the system along the trajectory. In this process, the time-varying transition rate of the semi-Markov process is introduced, and the mathematical expectation of the time-varying transition rate is used to handle the stochastic jump term, resulting in a derivative expression containing the system state vector and error term. The H∞ performance index and external disturbance vector are introduced to construct a dissipative inequality containing disturbance suppression performance. The state equation of the closed-loop control system is substituted into the dissipative inequality, and the results are rearranged to obtain sufficient conditions for the closed-loop control system to satisfy stochastic stability and possess H∞ performance. The sufficient conditions are transformed into matrix inequality form, which includes the sequence of positive definite matrices, the system matrix corresponding to the system modes, the input matrix, the disturbance matrix, the output matrix, the machine computation gain to be determined, and the preset human experience gain.

[0014] Preferably, the process of solving the matrix inequality criterion using linear matrix inequality techniques includes: The matrix inequalities are transformed by performing a congruence transformation, multiplying both sides of the inequality by a diagonal matrix composed of the inverses of a sequence of positive definite matrices, to eliminate nonlinear coupling terms. New decision variables are introduced for substitution: the first decision variable is defined as the inverse of the sequence of positive definite matrices, and the second decision variable is defined as the product of the machine computation gain and the first decision variable. This transforms the coupling terms in the original matrix inequalities into linear terms with respect to the new decision variables. Schur's complement lemma is then applied to reduce the order and linearize the processed matrix inequalities, constructing a set of standard-form linear matrix inequalities. These linear matrix inequalities include the first decision variable, the second decision variable, the human experience gain, the H∞ performance index, and the expected value of the transition rate.

[0015] Preferably, the process of deploying a complete human-machine collaborative asynchronous controller to an unmanned vessel includes: Real-time acquisition of motion state data and environmental parameters of unmanned ships, calculation of the current moment's premise variable values ​​and the membership degree of each fuzzy rule; The system identifies the modal information currently received by the controller and calls the corresponding machine-calculated gain and human-experienced gain from the memory; it calculates the algebraic sum of the two gains and performs weighted aggregation by combining real-time status data and membership degrees to generate the final control command, which is then sent to the ship's propulsion and steering gear system.

[0016] A second aspect of this invention provides a UMV human-machine cooperative asynchronous H∞ control system based on fuzzy Semi-Markov theory modeling, comprising: The TS fuzzy Semi-Markov modeling module is used to construct TS fuzzy Semi-Markov system models that can describe the nonlinear dynamic characteristics of unmanned ships and determine the semi-Markov process transition rate matrix of the system modes. The human-machine collaborative structure definition module is used to set the structure of the human-machine collaborative asynchronous controller based on the TS fuzzy Semi-Markov system model, define the asynchronous mapping relationship between the controller mode and the system mode, and inject the preset human experience gain into the structure to construct a controller framework containing the machine computing gain to be determined. The gain optimization solution module is used to receive the controller framework, construct a matrix inequality criterion that enables the closed-loop control system to satisfy the stochastic stability and H∞ performance index, solve the numerical solution of the machine-calculated gain using linear matrix inequality techniques, and combine the numerical solution with the human experience gain to generate a complete human-machine collaborative asynchronous controller. The real-time status monitoring module is used to collect system status vectors in real time during the operation of unmanned vessels, calculate the membership degree at the current moment according to the definition of fuzzy rules, identify the current controller modal index, and transmit the monitoring data to the cooperative control module. The asynchronous collaborative control module is used to call the corresponding machine calculation gain and human experience gain in the complete human-machine collaborative asynchronous controller according to the data output by the real-time status monitoring module, superimpose and weight the two gains, calculate the final control input signal and send it to the actuator of the unmanned vessel.

[0017] This invention provides a UMV human-machine cooperative asynchronous H∞ control method based on fuzzy Semi-Markov theory modeling. It has the following beneficial effects: 1. This invention constructs a TS fuzzy Semi-Markov system model, uses fuzzy logic to approximate the nonlinear dynamic characteristics of unmanned vessels, and employs a semi-Markov process to describe the stochastic switching between system modes. This method removes the constraint that the dwell time must follow an exponential distribution in traditional Markov jump system models, enabling a more accurate characterization of the dynamic behavior of unmanned vessels under complex sea conditions and variable load conditions, thereby improving the fitting accuracy of the control system design model to the actual physical process.

[0018] 2. This invention establishes an asynchronous mapping relationship between controller modes and system modes, allowing for inconsistencies between the modal information received by the controller and the current actual physical modes of the system. This asynchronous control mechanism effectively solves the modal mismatch problem caused by sensor measurement delays, signal transmission lags, or data packet loss, ensuring that the closed-loop control system maintains stochastic stability and meets the H∞ disturbance suppression performance index even in environments with communication network constraints or signal interference, thus enhancing the robustness of the system.

[0019] 3. This invention employs a human-machine collaborative control gain structure, superimposing the machine-calculated gain to be determined with a preset human experience gain. This method injects expert knowledge or historical experience data on ship maneuvering as prior constraints into the controller design. It utilizes the optimization computational power of linear matrix inequality techniques while retaining maneuvering characteristics consistent with maritime practice. Thus, while ensuring control accuracy, it avoids unreasonable control responses resulting from relying solely on numerical calculations, thereby improving the practicality of the control strategy. Attached Figure Description

[0020] Figure 1 This is a schematic diagram of the method flow of the present invention; Figure 2 This is a system architecture diagram of the present invention; Figure 3 This is the modal asynchronous diagram of the present invention; Figure 4 This is an open-loop diagram of the UMV system of the present invention; Figure 5 This is a closed-loop diagram of the UMV system of the present invention; Figure 6 This is a diagram of the control signals for the UMV system of the present invention. Detailed Implementation

[0021] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0022] Reference Figures 1-6 , Figure 1 This is a flowchart of a UMV human-machine cooperative asynchronous H∞ control method based on fuzzy Semi-Markov theory modeling, according to an embodiment of the present invention. The present invention provides a UMV human-machine cooperative asynchronous H∞ control method based on fuzzy Semi-Markov theory modeling, which may include: S100. Construct a TS fuzzy semi-Markov system model for an unmanned vessel (UMV). This model uses fuzzy logic to handle the nonlinear dynamic characteristics of the UMV, and employs a semi-Markov process to describe the stochastic switching characteristics between system modes. The system mode index is generated by the semi-Markov process. The model includes system matrices, input matrices, disturbance matrices, and output matrices corresponding to each mode. These matrices are all known constant matrices with consistent dimensions. By setting the number of fuzzy rules and defining the membership function for each rule, the weighted summation of the local linear state-space equations established based on different fuzzy rules is performed using the membership function to obtain the global TS fuzzy semi-Markov system model.

[0023] S200, based on the constructed TS fuzzy semi-Markov system model, designs the structure of a human-machine collaborative asynchronous controller. The controller employs a parallel distributed compensation strategy, sharing the same preconditions and membership functions as the system model. The total control gain of the controller is defined as the superposition of the undetermined machine computation gain and a preset human experience gain. Controller modal indices and system modal indices are defined, allowing for inconsistencies between their values, thereby constructing an asynchronous mapping relationship between the controller and system modes. Both the machine computation gain and the human experience gain depend on the controller modal index and the fuzzy rule index.

[0024] S300: Substituting the designed human-machine cooperative asynchronous controller into the TS fuzzy Semi-Markov system model, a closed-loop control system is obtained. A stochastic Lyapunov function dependent on the system mode index is constructed. This function is set as a quadratic form of the system state vector and a sequence of positive definite matrices to be determined, with the positive definite matrix sequence taking different values ​​as the system mode index changes. Weak infinitesimal operator operations are performed on the stochastic Lyapunov function, introducing the time-varying transition rate of the semi-Markov process. The mathematical expectation of the time-varying transition rate is used to handle the stochastic jump term, and the differential expectation of the system along the trajectory is calculated. Introducing the H∞ performance index and the external disturbance vector, a dissipative inequality including disturbance suppression performance is constructed. Substituting the state equation of the closed-loop control system into the dissipative inequality, and rearranging, a sufficient condition is obtained for the closed-loop control system to satisfy stochastic stability and possess H∞ performance. This sufficient condition is then transformed into a matrix inequality form.

[0025] S400 utilizes linear matrix inequality techniques to solve the obtained matrix inequality criterion. A congruent transformation is performed on the matrix inequality by multiplying both sides of the inequality by a diagonal matrix composed of the inverses of a sequence of positive definite matrices, eliminating nonlinear coupling terms. New decision variables are introduced: the first decision variable is defined as the inverse of the sequence of positive definite matrices, and the second decision variable is defined as the product of the machine computational gain and the first decision variable. This transforms the coupling terms in the original matrix inequality into linear terms with respect to the new decision variables. Schur's complement lemma is applied to reduce the order and linearize the processed matrix inequalities, constructing a set of standard-form linear matrix inequalities. Solving these linear matrix inequalities yields a numerical solution for the machine computational gain. This numerical solution is then combined with a pre-defined human experience gain to generate a complete human-machine collaborative asynchronous controller.

[0026] The S500 deploys a complete human-machine collaborative asynchronous controller to unmanned vessels. During operation, it collects real-time motion state data and environmental parameters, calculates the current context variable values ​​and membership degrees of various fuzzy rules, identifies the modal information received by the controller, and retrieves the corresponding machine-calculated gain and human-experienced gain from memory. It calculates the algebraic sum of these two gains, combines them with real-time state data and membership degrees for weighted aggregation, and generates the final control command, which is sent to the vessel's propulsion and steering systems to achieve navigation control of the unmanned vessel.

[0027] Reference Figures 1-6 S100, Construct a TS fuzzy Semi-Markov system model for the unmanned UMV. This step S100 specifically includes: First, the nonlinear dynamic equations of the unmanned vessel are established in both the body coordinate system and the Earth-fixed coordinate system. The system state vector of the unmanned vessel is then defined. This vector contains the ship's displacement and velocity information in the three degrees of freedom of pitch, sway, and yaw. Specifically, the velocity vector is defined as follows: Represent the sway velocity, roll velocity, and pitch angular velocity in volume coordinates; define the position vector. It represents the X-axis position, Y-axis position, and bow angle in the Earth's fixed coordinate system.

[0028] Considering that unmanned vessels will face different load conditions (e.g., no load, full load) or changes in immersion depth during actual operations, these changes in physical characteristics will cause jumps in the system's inertia matrix and damping matrix. Therefore, a semi-Markov process is introduced. To describe the random switching between these operating modes. When hour( , (Total number of modes), the dynamic equations of the unmanned vessel are described as follows: ; In the formula, The time derivative of the ship's velocity vector, i.e., the acceleration vector; This represents the velocity vector of the unmanned vessel. This represents the position or pose vector of an unmanned vessel. The inertia matrix includes the added mass; It is a damping matrix that includes Coriolis force, centripetal force, and hydrodynamic damping; The restoring force matrix (mainly derived from anchoring force or buoyancy); Configure the matrix for control inputs; Configure a matrix for external interference; It is the control torque input vector; It is a vector simulating external disturbances such as wind, waves, and currents.

[0029] To address the nonlinear characteristics present in the above dynamic equations (mainly concentrated in the damping matrix) (The coupling terms in the model), this embodiment uses Takagi-Sugeno (TS) fuzzy modeling technology for global approximation. Prerequisite variables are selected. and (Usually chosen as the velocity component), the number of fuzzy rules is set to... and for each fuzzy rule ( Define the corresponding fuzzy set membership function. .

[0030] For each system mode determined by a semi-Markov process , establish the first The local linear state-space equations corresponding to the fuzzy rules are: rule If the prerequisite variable Belongs to fuzzy set ,and Belongs to fuzzy set The system state equation is then described as follows: ; The system's performance output equation is defined as follows: ; In the formula, The time derivative of the system state vector, i.e., the state rate of change vector; This represents the system state vector, which typically contains displacement and velocity information of the unmanned vessel in the pitch, sway, and yaw directions. This represents the system's performance output vector; It is the control torque input vector; It is a vector simulating external disturbances such as wind, waves, and currents; , , , , These correspond to the modes. and fuzzy rules The system matrix, input matrix, perturbation matrix, output matrix, and perturbation-output coupling matrix are defined. These matrices are obtained by applying nonlinear matrices at specific operating points. , The constant matrix is ​​obtained by linearizing and recombining the linearized matrix.

[0031] Using normalized membership functions By weighted summing of all local linear state-space equations, a global TS fuzzy Semi-Markov system model describing the nonlinear dynamic characteristics of unmanned vessels is obtained: ; ; In the formula, The time derivative of the system state vector, i.e., the state rate of change vector; This represents the system's performance output vector; , , , , These correspond to the modes. and fuzzy rules The system matrix, input matrix, perturbation matrix, output matrix, and perturbation-output coupling matrix are defined. These matrices are obtained by applying nonlinear matrices at specific operating points. This represents the system state vector, which typically contains displacement and velocity information of the unmanned vessel in the pitch, sway, and yaw directions. It is the control torque input vector; It is a vector simulating external disturbances such as wind, waves, and currents; Represents the total number of fuzzy rules; An index representing a fuzzy rule; Indicates the first Normalized membership function of a fuzzy rule.

[0032] Finally, a specific definition of the semi-Markov process describing system mode switching is provided. Unlike traditional Markov processes that assume residence time follows an exponential distribution (i.e., memoryless), this embodiment uses a probability density function. To describe system modes Duration of stay The dwell time follows a non-exponential distribution, preferably a Weibull distribution. Let... For shape parameters, If is the scale parameter, then its probability density function is expressed as: ; In the formula, This indicates that the system is in a modal state. The probability density function of the dwell time; This represents the shape parameter of the Weibull distribution; The scale parameter representing the Weibull distribution; Indicates the length of stay; This represents an exponential function.

[0033] Under this description, from modality Jump to modality transfer rate It is no longer a constant, but depends on the current dwell time. The time-varying transfer rate function is used. To facilitate subsequent engineering solutions for the controller, a constantized transfer rate matrix is ​​obtained by calculating the mathematical expectation of the time-varying transfer rate function. This approach preserves the ability of semi-Markov processes to accurately describe the distribution of state durations under complex sea states, while also allowing the model parameters to be adapted to the solution framework of linear matrix inequalities.

[0034] Reference Figures 1-6 S200: Based on the TS fuzzy Semi-Markov system model, design the structure of a human-machine collaborative asynchronous controller. This step S200 specifically includes: To effectively compensate the nonlinear TS fuzzy system established in step S100, a parallel distributed compensation strategy is adopted to construct the controller framework. This means that the fuzzy rule structure of the controller shares the exact same fuzzy premise variables as the system model of the unmanned vessel. Number of fuzzy rules and membership function For each fuzzy rule, design a corresponding local linear control law.

[0035] Regarding the composition of the control gain, an innovative superposition structure is adopted to combine the control gain matrix. Decomposed into two parts: machine calculation of the gain matrix and human experience gain matrix Among them, subscript Indicates the fuzzy rule index, subscript This indicates the controller modal index. The specific control input... Expressed as a weighted sum of the local control laws: ; In the formula, Represents the control input vector; Represents the total number of fuzzy rules; Indicates a fuzzy rule index; Indicates the first The normalized membership function of the rule; This indicates that the machine calculates the gain matrix; Represents the human experience gain matrix; This represents the system state vector.

[0036] In this structure, the human experience gain matrix This is a known constant matrix pre-injected into the controller. The values ​​of this matrix are obtained from historical maneuvering data of the unmanned vessel under specific sea conditions or from an expert knowledge base. For example, in collision avoidance or navigation in rough seas, experienced captains typically employ specific steering amplitude limits or propeller response biases, quantifying these experiences into fixed gain values ​​and constructing them into a matrix. This serves as prior knowledge to constrain or guide control behavior. The advantage of this approach is that it avoids the potentially aggressive or impractical control commands that pure numerical optimization might produce. Meanwhile, the machine calculates the gain matrix. These are the decision variables to be optimized, and their specific values ​​are determined through the subsequent linear matrix inequality solution process to ensure that the system meets mathematical stability and performance indicators.

[0037] Define the asynchronous mapping relationship between controller modes and system modes. In practical networked control systems for unmanned ships, due to the inevitable time consumption in sensor acquisition, signal encoding, wireless channel transmission, and decoding processes, the modal information received by the controller often lags behind the current actual physical mode of the system.

[0038] set up ( This represents the current true mode of the system (determined by the Semi-Markov process), and is set as follows: ( This indicates the current mode of the controller. The asynchronous description mechanism constructed in this embodiment allows... This refers to the inconsistency between the controller modal index and the system modal index. This inconsistency characterizes modal mismatch caused by sensor measurement delays, signal transmission lags, or data packet loss. Mathematically, this asynchronicity can be addressed by introducing conditional probabilities or by directly using different modal indices in the system matrix and control matrix, thus freeing controller design from the stringent assumption that "the system modes must be known accurately in real time."

[0039] Finally, substituting the aforementioned human-machine collaborative asynchronous controller into the TS fuzzy Semi-Markov system model constructed in step one, the dynamic equations of the closed-loop control system are obtained. By merging the system state equations and the control input equations, the closed-loop system is described as follows: ; ; In the formula, The time derivative of the system state vector, i.e., the state rate of change vector; This represents the system's performance output vector; , , , , These correspond to the modes. and fuzzy rules The system matrix, input matrix, perturbation matrix, output matrix, and perturbation-output coupling matrix are defined. These matrices are obtained by applying nonlinear matrices at specific operating points. Represents the total number of fuzzy rules; An index representing a fuzzy rule; Indicates the first Normalized membership function of a fuzzy rule; It is a vector simulating external disturbances such as wind, waves, and currents; This indicates that the machine calculates the gain matrix; Represents the human experience gain matrix; Represents the system state vector; Indicates the first Normalized membership function of fuzzy rules for bar controllers; This represents the fuzzy rule index of the controller.

[0040] The closed-loop equation clearly demonstrates that the system matrix depends on the system modes. The control gain depends on the controller mode. Its asynchronous nature also reflects the gain of human experience. It participates directly in the pole placement of the closed-loop system as a fixed bias term. Subsequent stability analysis will be based on this closed-loop equation.

[0041] Reference Figures 1-6 Step S300 involves constructing a stochastic Lyapunov function and establishing stochastic stability and H∞ performance criteria for the UMV system. This step S300 specifically includes: For the TS fuzzy Semi-Markov closed-loop system of unmanned ships, a stochastic Lyapunov functional dependent on the system modes is constructed. Unlike traditional public Lyapunov functions, the function form selected in this embodiment can reflect the current modal characteristics of the system in real time. Specifically, it is expressed as a quadratic form of the system state vector and a sequence of positive definite matrices: ; In the formula, Let this be the system's state vector; Represents a mode-dependent stochastic Lyapunov functional; This represents the transpose of the state vector; Represents a sequence of positive definite symmetric matrices that depend on modes; Represents a semi-Markov random process; Indicates the current system modality index ( ). It is a set of positive definite symmetric matrices to be determined when the system is in mode. hour, Values This matrix not only characterizes the generalized energy of the system, but its modal switching properties also encompass its adaptability to modal transitions.

[0042] The weak infinitesimal operator operation is performed on the Lyapunov function mentioned above, denoted as . Because this invention employs a semi-Markov process instead of a conventional Markov process, the mode transition rate is higher. It's about the length of stay. The time-varying function. To handle this complex stochastic process, step S300 introduces a method for processing mathematical expectation when calculating the operator. Specifically, The calculation consists of two parts: The derivative term of the continuous trajectory: the differential term along the trajectory of the system's state equation, expressed as... Substituting the closed-loop system equations obtained in step S200 into this equation, which includes the asynchronous controller gain, we arrive at the following: .

[0043] Random jump term: manifested as the current mode Jump to all possible target modes The probability-weighted sum. This embodiment utilizes the mathematical expectation of the transition rate. To replace the time-varying transfer rate, thus transforming complex integral and differential terms into linear algebraic terms, i.e. .

[0044] To endow the system with the ability to resist wind and wave interference, the H∞ performance index is introduced. Define a concept related to performance output. and external disturbances Dissipation inequality: ; In the formula, Represents the mathematical expectation operator; Denotes a weak infinitesimal operator for a stochastic Lyapunov functional; This represents the energy output of the performance; This represents the H∞ performance index; This represents the energy input from external disturbances.

[0045] The physical meaning of this inequality is: the rate of change of system energy plus the energy of the output error should be less than the energy of the external disturbance. If this inequality holds, it indicates that the system has strict dissipation, meaning that the impact of external disturbances on the system's performance output is limited to a factor of 1.5. Below the horizontal level.

[0046] By applying the generalized Dynkin formula, the above differential inequality is transformed into an integral form to verify the stochastic stability over long time scales.

[0047] By Substituting the specific expansion into the above dissipation inequality and rearranging like terms, we can derive a matrix containing the system matrix. Input matrix Asynchronous control gain Lyapunov matrix and expected transfer rate Sufficient and necessary conditions for matrix inequalities.

[0048] The following technical features are clearly embodied in this matrix inequality: The effect of asynchronous mapping: the index of the control term used as the left and right multiplication factor in the inequality is... The system core matrix index is By traversing all possible Combining (using conditional probability or total probability formulas to cover) ensures that the energy function converges (i.e., the derivative is negative) under any asynchronous misalignment.

[0049] The manifestation of human-machine collaboration: the control gain term in the inequality explicitly preserves the constant matrix. (Human experience). This means that the final stability condition is derived under the premise of "known human intervention," thus ensuring the gain calculated by the machine. It can complement human experience gains and work together to maintain the stability of the closed-loop system.

[0050] Finally, the resulting matrix inequality is a nonlinear matrix inequality (BMI) because it includes decision variables. and the gain to be determined The product term cannot be directly solved using conventional convex optimization algorithms. This provides the theoretical input and mathematical foundation for introducing congruent transformation and variable substitution in subsequent step four. The output of this step is a set of primitive algebraic criteria that can determine whether a closed-loop UMV system is stochastically stable and satisfies H∞ performance.

[0051] Reference Figures 1-6 Step S400 involves solving the aforementioned stability criterion using the Linear Matrix Inequality (LMI) technique to obtain the machine computation gain matrix and generate the final human-machine collaborative asynchronous controller. Step S400 specifically includes: Convex optimization is performed on the stochastic stability and H∞ performance criteria of the semi-Markov jump system derived in step S300. This is because the original matrix inequalities contain Lyapunov matrices. With the controller gain to be determined The product terms constitute nonlinear coupling, making it impossible to solve directly using conventional convex optimization algorithms. Therefore, a congruent transformation strategy is introduced. The transformation matrix is ​​defined as a block diagonal matrix, whose diagonal elements are mainly composed of the inverse of the Lyapunov matrix. The transformation matrix is ​​then multiplied on the left and right sides of the original matrix inequality to perform an equivalent transformation on the inequality, thereby eliminating the nonlinear coupling structure.

[0052] Perform variable substitution to linearize the inequalities. Introduce new decision variables. The relationship between it and the machine computation gain and the inverse of the system state matrix is ​​defined as follows: During this replacement process, the human experience gain preset in the controller is... Since it is a known constant matrix (derived from historical data or expert experience), it is directly expressed in the transformed inequality as follows: In the form of the decision variable. The linear relationship is preserved, therefore no additional variable substitution is required. Through this process, the nonlinear term in the original criterion is eliminated. Transformed into a new variable and linear combination terms This effectively transforms the controller design problem into a convex optimization problem, while preserving the constraint of human experience gain on the pole distribution of the closed-loop system.

[0053] By applying the Schur Complement Lemma, inequalities containing quadratic form terms are reduced in order, transforming nonlinear inequalities into standard linear matrix inequalities (LMI) form. Specifically, the following block matrix inequalities are constructed: For all possible system modes and controller mode and all combinations of fuzzy rules It must meet the following requirements: ; In the formula, the symbol This represents the symmetric transpose of the matrix; the physical and mathematical meanings of each matrix block element are defined as follows: Corresponding to the active dynamic characteristics of the system state, its specific form is as follows: ; In the formula, Represents the system dynamic matrix; Represents the control input matrix; Denotes the inverse of the Lyapunov matrix; Represents the intermediate decision variable matrix; Represents the human experience gain matrix; This indicates that the system remains in the current mode. The expected value of the self-transfer rate; This represents the Hermitian operator, and this block embodies the system's internal dynamic equilibrium and semi-Markov transition rate. Impact on stability; Corresponding to the perturbation input channel, it contains the perturbation matrix. With disturbance-output coupling term; and Performance metrics related to H∞ on the diagonal And the negative definite terms of the identity matrix, used to constrain external disturbances. Performance output The energy gain is less than the performance index. .

[0054] The set of inequalities ensures that the derivative of the energy function of the closed-loop system is always negative (i.e., energy dissipation) under any asynchronous mode combination and fuzzy rule activation state, thus guaranteeing the stochastic stability of the system.

[0055] Finally, the linear matrix inequality toolbox in the mathematical computing software was used to numerically solve the above system of linear matrix inequalities. The solution process aimed to find a symmetric positive definite matrix that satisfies all constraints. and general matrix variables After obtaining such feasible solutions, matrix inversion is performed. Recover the gain matrix required for machine computation. (The calculated...) Compared with the human experience gain preset in step S200 By performing linear superposition, the final human-machine cooperative asynchronous controller gain is obtained. The gain matrix parameters are then programmed or configured into the control system of the unmanned vessel, completing the construction and deployment of the controller. In this way, the controller combines the robustness of mathematical optimization with the driving expertise of human specialists.

[0056] Reference Figures 1-6 S500, simulation verification and parameter setting. This step S500 specifically includes: Initialize the UMV system model parameters and multimodal environment settings. Select an unmanned vessel with typical nonlinear characteristics as the controlled object, and establish a dynamic model in three degrees of freedom: pitch, sway, and yaw. To simulate the impact of changes in ship load (e.g., no load, full load, half load) and draft on hydrodynamic coefficients under actual sea conditions, set the total number of system modes. Different inertia matrix, damping matrix, and restoring force matrix are defined for these two modes respectively.

[0057] For the Semi-Markov process describing mode switching, unlike the exponential distribution assumption of traditional Markov chains, this embodiment assumes that the residence time of each mode follows a Weibull distribution. Specifically, shape parameters are set... and scale parameters For a specific value (e.g.) This method simulates sea state changes, which exhibit certain memory characteristics and non-constant failure rates. Based on the generalized probability density function, the expected transition rate matrix between each mode is calculated for subsequent controller gain calculation.

[0058] Configure the TS fuzzy logic and human-machine collaborative controller parameters. Select the ship's pitch and sway speeds as fuzzy premise variables and set the number of fuzzy rules. A Gaussian membership function is used to partition the workspace to ensure that the global model can smoothly approximate the nonlinear dynamic characteristics of UMV.

[0059] In the controller design phase, a human-machine collaboration strategy is implemented.

[0060] Human experience injection: Based on the piloting experience of senior captains, a fixed set of experience gain matrices is set. The matrix has a relatively small and conservative value, which is intended to provide basic damping for the system and prevent overly aggressive control actions due to the pursuit of mathematical optimality.

[0061] Machine calculation of gain: Introducing the H∞ performance index To characterize the suppression capability of external wind and wave disturbances, γ is used as an undetermined parameter in the linear matrix inequality. Using the LMI toolbox in Matlab (such as the YALMIP solver), the linear matrix inequality criterion derived in step S400 is input. Under the premise of satisfying the stochastic stability constraint, the optimal machine computation gain matrix is ​​solved. The final controller parameters are the sum of the gains from the two parts mentioned above.

[0062] Then, asynchronous control simulation and result analysis were performed. During the simulation, random wind and wave disturbances simulating real sea conditions were applied. .

[0063] Reference Figures 1-6 , Figure 2 This is a diagram of the architecture of a UMV human-machine collaborative asynchronous H∞ control system according to an embodiment of the present invention.

[0064] To implement the aforementioned control method, this invention also provides a UMV human-machine collaborative asynchronous H∞ control system based on fuzzy Semi-Markov theory modeling. This system mainly consists of five core functional modules, which interact with each other via a data bus to collaboratively complete complex control tasks.

[0065] TS Fuzzy Semi-Markov Modeling Module: This module is the foundation of the entire control system, and its main function is to establish a digital mathematical model of the unmanned vessel.

[0066] Specifically, the TS fuzzy semi-Markov modeling module internally stores the physical parameters (such as the mass matrix and damping coefficient matrix) of the unmanned vessel under different operating conditions, including unloaded, fully loaded, and partially loaded conditions. Based on the definitions in step S100, it uses non-exponential probability density functions such as the Weibull distribution to describe the system modes and their dwell time characteristics, and calculates the mathematical expectation matrix of the time-varying transition rate. Simultaneously, this TS fuzzy semi-Markov modeling module is responsible for running fuzzy logic algorithms to decompose the nonlinear dynamic characteristics of the vessel into multiple locally linear TS fuzzy subsystems, providing accurate "system matrix" and "input matrix" parameter streams for subsequent controller design.

[0067] Human-Machine Collaboration Structure Definition Module: This module is responsible for building the "skeleton" of the controller.

[0068] Following the strategy in step S200, the parallel distributed compensation (PDC) principle is first adopted to ensure that the controller's fuzzy rules are consistent with those of the modeling module. Secondly, the human-machine collaborative structure definition module incorporates an "expert experience base," storing fixed gain matrices summarized from experienced captains or historical data. Most importantly, this human-machine collaborative structure definition module defines an "asynchronous mapping table," which clarifies the controller modal index when sensor delays or packet loss occur. With system physical mode index The correspondence or non-correspondence between them allows the system to maintain the integrity of the control architecture even in the event of modal mismatch.

[0069] Gain optimization solution module: This module is the "computational brain" of the system and usually runs during the system initialization phase or offline phase.

[0070] Receive model parameters and structure definitions from the TS fuzzy Semi-Markov modeling module and the human-machine collaborative structure definition module, including the H∞ anti-interference performance index. Under the constraints of stochastic Lyapunov stability, a system of high-dimensional linear matrix inequalities is automatically constructed. This gain optimization module calls an embedded numerical optimization solver (such as the Interior-point method solver) to iteratively solve these inequalities and calculate the optimal machine-computed gain matrix. After the calculation is complete, the module will... and Logical binding is performed to generate a complete gain lookup table matrix, which is then stored in the controller's memory.

[0071] Real-time status monitoring module: This module is the interface connecting the physical world and digital algorithms.

[0072] During ship operation, this real-time status monitoring module uses sensors such as IMU, GPS, and current meters to collect state vectors such as pitch displacement, sway velocity, and bow roll rate in real time at millisecond intervals. .

[0073] Asynchronous Cooperative Control Module: This module directly drives the output of the actuator.

[0074] Based on the "membership degree" and "controller modal index" provided by the real-time status monitoring module, it retrieves the corresponding machine-calculated gain from the memory. and human experience gain .

[0075] The asynchronous cooperative control module performs core algebraic operations: first, it linearly superimposes the two gains to obtain the total gain; then, it uses membership degrees to perform a weighted summation of the local control laws under each rule. The final calculated control commands are sent to the propulsion and steering systems of the unmanned vessel via a D / A converter or CAN bus, driving the vessel to overcome wind and wave interference and achieve stable navigation along a predetermined trajectory.

[0076] Through the cooperation of the above five modules, the system realizes a complete closed loop from mathematical modeling and offline optimization to online asynchronous control, effectively solving the technical problems that traditional control systems struggle to balance random mode switching, signal transmission delay, and the integration of human experience.

Claims

1. A UMV human-machine collaborative asynchronous H∞ control method based on fuzzy Semi-Markov theory modeling, characterized in that, The method comprises the following steps: S1, constructing a T-S fuzzy Semi-Markov system model of an unmanned marine vehicle (UMV), wherein the system model uses fuzzy logic to approximate the nonlinear dynamic characteristics of the unmanned marine vehicle, and uses a semi-Markov process to describe the random switching characteristics between system modes; S2, designing a structure of a human-machine collaborative asynchronous controller based on the T-S fuzzy Semi-Markov system model, setting a total control gain of the asynchronous controller to be a superposition of a machine calculation gain to be determined and a preset human experience gain, and defining an asynchronous mapping relationship between a controller mode and a system mode; S3, substituting the human-machine collaborative asynchronous controller into the T-S fuzzy Semi-Markov system model to obtain a closed-loop control system, constructing a Lyapunov function dependent on the system mode, and deriving a matrix inequality criterion for making the closed-loop control system satisfy random stability and an H performance index; S4, solving the matrix inequality criterion by using a linear matrix inequality technique, obtaining a numerical solution of the machine calculation gain, and combining the numerical solution with the human experience gain to generate a complete human-machine collaborative asynchronous controller; S5, deploying the complete human-machine collaborative asynchronous controller to the unmanned marine vehicle, calculating a control input according to real-time collected state data, and performing navigation control on the unmanned marine vehicle. 2.The UMV human-machine collaborative asynchronous H∞ control method based on fuzzy Semi-Markov theory modeling of claim 1, wherein, In step S1, the T-S fuzzy Semi-Markov system model of the unmanned marine vehicle (UMV) comprises the following steps: defining a system state vector, a control input signal vector, an external disturbance vector, and a performance output vector of the unmanned marine vehicle, wherein the system state vector comprises displacement and velocity information of the marine vehicle in surge, sway, and yaw directions; setting the number of fuzzy rules, and defining corresponding premise variables and membership functions for each fuzzy rule; for each system mode determined by a semi-Markov process, establishing a local linear state space equation corresponding to the fuzzy rule, wherein the local linear state space equation comprises a system matrix, an input matrix, and a disturbance matrix related to the current system mode; performing weighted summation on all local linear state space equations by using the membership functions, thereby obtaining a global T-S fuzzy Semi-Markov system model capable of describing nonlinear dynamic characteristics of the unmanned marine vehicle under different operating conditions. 3.The UMV human-machine collaborative asynchronous H∞ control method based on fuzzy Semi-Markov theory modeling of claim 1, wherein, In step S1, the semi-Markov process is used to describe the random switching characteristics between system modes, which comprises the following steps: using a probability density function to describe the residence time distribution of the system mode in a certain state, wherein the residence time distribution is subject to a non-exponential distribution, so that the transition rate between modes becomes a time-varying function dependent on the residence time; calculating the mathematical expectation of the time-varying transition rate function to obtain a constant transition rate matrix; introducing the system mode index as a variable into the related matrix parameters of the system model, so that the dynamic parameters of the system model jump with the random switching of the system mode, thereby representing the dynamic characteristic changes of the unmanned marine vehicle under different loads or sea environments. 4.The UMV human-machine collaborative asynchronous H∞ control method based on fuzzy Semi-Markov theory modeling of claim 1, wherein, In step S2, the structure of the designed human-robot collaborative asynchronous controller specifically includes: The fuzzy rule structure of the controller is designed using a parallel distributed compensation strategy, so that it shares the same premise variables and membership functions as the T-S fuzzy Semi-Markov system model; The form of the local control law of the controller is defined, and each local control law corresponds to a fuzzy rule; The gain matrix structure of the local control law is set to be in an additive combination form, that is, it is directly composed by adding a machine-calculated gain matrix to be solved and a pre-set human-experience gain matrix; The local control law containing the superimposed gain is weighted and summed through the membership function to obtain the final control input expression.

5. The UMV human-machine collaborative asynchronous H∞ control method based on fuzzy Semi-Markov theory modeling according to claim 1, characterized in that, In step S2, the asynchronous mapping relationship between the controller mode and the system mode specifically includes: The system mode index is set to represent the current actual physical mode of the unmanned ship, and the controller mode index is set to represent the mode information received at the controller end; An asynchronous description mechanism is constructed to allow the controller mode index and the system mode index to take different values, thereby representing the mode mismatch phenomenon between the controller and the system caused by sensor measurement delay, signal transmission time lag or data packet loss; In the controller design, both the machine-calculated gain and the human-experience gain depend on the controller mode index and the fuzzy rule index, rather than directly on the system mode index. 6.The UMV human-machine collaborative asynchronous H∞ control method based on fuzzy Semi-Markov theory modeling of claim 1, wherein, In step S2, the human-experience gain is obtained in the following manner: According to the historical operation data or expert knowledge base of the unmanned ship in a specific sea state, fixed gain values under different controller modes and fuzzy rules are determined; The fixed gain values are constructed as a known constant matrix, which is injected into the controller structure as prior knowledge to constrain or guide the solution direction of the machine-calculated gain during the optimization process; The machine-calculated gain is used as a decision variable to be optimized, and its specific value is determined through a subsequent linear matrix inequality solving process.

7. The UMV human-machine collaborative asynchronous H∞ control method based on fuzzy Semi-Markov theory modeling according to claim 1, characterized in that, In step S3, the matrix inequality criterion that makes the closed-loop control system satisfy the random stability and H performance index specifically includes: A random Lyapunov function dependent on the system mode index is constructed, which is set to be in the form of a quadratic function of the system state vector and a to-be-determined positive definite matrix sequence, and the positive definite matrix sequence takes different values with the change of the system mode index; The weak infinitesimal operator operation is performed on the random Lyapunov function to calculate the differential expectation of the system along the trajectory, in which process the time-varying transition rate of the semi-Markov process is introduced, and the mathematical expectation of the time-varying transition rate is used to process the random jump term to obtain a derivative expression containing the system state vector and an error term; The preset H performance index and an external disturbance vector are introduced to construct a dissipation inequality containing disturbance suppression performance, and the state equation of the closed-loop control system is substituted into the dissipation inequality to obtain a sufficient condition for the closed-loop control system to satisfy the random stability and have the H performance. The sufficient condition is converted into a matrix inequality form, and the matrix inequality includes the sequence of positive definite matrices, system matrices corresponding to system modes, an input matrix, a disturbance matrix, an output matrix, a machine calculation gain to be determined, and a preset human experience gain. 8.The UMV human-machine collaborative asynchronous H∞ control method based on fuzzy Semi-Markov theory modeling of claim 1, wherein, In step S4, the solving of the matrix inequality criterion by using the linear matrix inequality technique specifically includes: Performing a contract transformation on the matrix inequality, and multiplying a diagonal matrix composed of inverse matrices of the sequence of positive definite matrices on the left and right sides of the inequality respectively, to eliminate nonlinear coupling terms in the inequality; Introducing new decision variables for variable substitution, defining a first decision variable as an inverse matrix of the sequence of positive definite matrices, and defining a second decision variable as a product of the machine calculation gain and the first decision variable, so as to convert the coupling terms in the original matrix inequality into linear terms with respect to the new decision variables; Applying a Schur complement lemma to perform order reduction and linearization processing on the processed matrix inequality, and constructing a group of linear matrix inequalities in a standard form, which include the first decision variable, the second decision variable, the human experience gain, an H performance index, and a mathematical expectation of a transition rate. 9.The UMV human-machine collaborative asynchronous H∞ control method based on fuzzy Semi-Markov theory modeling of claim 1, wherein, In step S5, the deployment of the complete man-machine collaborative asynchronous controller to the unmanned ship specifically includes: Real-time acquisition of motion state data and environmental parameters of the unmanned ship, calculation of premise variable values at the current time and membership degrees of each fuzzy rule; Identification of modal information currently received by the controller, and calling of corresponding machine calculation gains and human experience gains in the memory; Calculation of an algebraic sum of the two parts of the gain, and weighted aggregation of the real-time state data and the membership degrees to generate a final control instruction which is sent to a ship propeller and a rudder system.

10. A UMV human-machine collaborative asynchronous H∞ control system based on fuzzy Semi-Markov theory modeling, applied to the UMV human-machine collaborative asynchronous H∞ control method based on fuzzy Semi-Markov theory modeling according to any one of claims 1-9, characterized in that, The system comprises: A T-S fuzzy Semi-Markov modeling module, which is configured to construct a T-S fuzzy Semi-Markov system model capable of describing nonlinear dynamic characteristics of the unmanned ship, and determine a semi-Markov process transition rate matrix of system modes; A man-machine collaborative structure defining module, which is configured to set a structure of the man-machine collaborative asynchronous controller based on the T-S fuzzy Semi-Markov system model, define an asynchronous mapping relationship between controller modes and system modes, and inject preset human experience gains into the structure to construct a controller framework including a machine calculation gain to be determined; A gain optimization solving module, which is configured to receive the controller framework, construct a matrix inequality criterion for making a closed-loop control system satisfy random stability and an H performance index, solve a numerical solution of the machine calculation gain by using a linear matrix inequality technique, and combine the numerical solution with the human experience gain to generate a complete man-machine collaborative asynchronous controller; A real-time state monitoring module, which is configured to, during operation of the unmanned ship, acquire a system state vector in real time, calculate a membership degree at the current time according to fuzzy rule definition, identify a current controller mode index, and transmit monitoring data to the collaborative control module. An asynchronous cooperative control module is configured to call corresponding machine calculation gain and human experience gain in the complete human-machine cooperative asynchronous controller according to the data output by the real-time state monitoring module, superimpose and weight aggregate the two parts of gain, calculate the final control input signal and send to the actuator of the unmanned ship.

Images