Ship LAAV motion model parameter identification method and system based on dynamic immune genetic algorithm

Abstract

The invention belongs to the technical field of ship motion, and discloses a ship LAAV motion model parameter identification method based on a dynamic immune genetic algorithm, which comprises the following steps of: constructing an identification model fitness criterion function by using ship real sea test data, setting a hyper-parameter adaptive adjustment mechanism, and performing several-generation selection, intersection, variation and immune operation on an initial population to identify ship LAAV motion model parameters. And finally determining model parameters. According to experimental verification, unknown parameters in a ship LAAV motion model are optimized by utilizing a DIGA framework in combination with sea test data, and an optimal solution is substituted into the LAAV model to generate AVP (Attribute, Velocity, Position) data in a simulation manner. The test result shows that the DIGA method has the advantages of high precision and high convergence speed in the LAAV model parameter identification problem, and the similarity between the generated AVP data and the sea test data is more than 90%.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of navigation and ship motion, and particularly relates to a ship LAAV motion model parameter identification method and system based on a hyperparameter natural adjustment dynamic immune genetic algorithm. BACKGROUND

[0002] A ship motion data generator is an important tool for navigation algorithm research, and its research background is driven by the dual needs of navigation simulation accuracy and intelligent shipping development. Its significance lies in generating high-precision trajectory data that meets the characteristics of real ships, improving navigation algorithm verification efficiency, supporting ship autonomous navigation and intelligent decision-making, promoting multidisciplinary technology integration, and providing data support for ocean economy and navigation safety.

[0003] Current international methods for generating ship motion data can be summarized as follows: (1) Typical trajectory segments (such as straight lines, turns, constant speed, variable speed climbing, and diving) are combined to obtain the motion trajectory of the carrier. This method is simple and easy to implement, and can simulate the motion of the carrier, but the generated motion trajectory is quite different from the actual maneuvering state, and does not meet the actual motion of the carrier. (2) Based on measured navigation data, relevant numerical algorithms are used to solve the inertial navigation equation. This trajectory generator design method has high precision performance and is close to the actual motion of the carrier; its disadvantage is that it requires a large amount of measured data as a basis, which is difficult to obtain in some cases. (3) A mathematical model is established based on the carrier dynamics equation or the steering equation. Zhang Bin combined a simple ship dynamics equation when designing the SINS trajectory generator, and analyzed the ship motion state to obtain the attitude and speed changes, and then solved the IMU output. The current ship hydrodynamic a&omega model generates a very close-to-real-ship-characteristics data generator, which uses the Levenberg-Marquardt algorithm to identify the unknown parameters in the a&omega model. The single initial value setting is easy to make the optimization result overflow, resulting in low identification accuracy. In addition, each identification can only input a single set of data for static identification, which has the defects of low identification efficiency and long time consumption.

[0004] In view of the problem of multiple identification parameters in the modeling process of the LAAV model, classical identification algorithms such as the least squares method and its derivatives only provide single-point estimation, which may lead to insufficient deterministic estimation of model parameters. Among the current common intelligent identification methods, there are three types of neural network algorithms, genetic algorithms, and particle swarm optimization algorithms. The neural network algorithm requires a large amount of training data, the particle swarm optimization algorithm has weak global search ability, and the genetic algorithm (GA) does not require a large amount of training samples for learning and weight setting, and has strong global optimal search ability.

[0005] Through the above analysis, the problems and defects of the prior art are:

[0006] In view of the problems of few training samples and many to-be-identified parameters in the LAAV model, the genetic algorithm is relatively suitable. However, the genetic algorithm itself also has certain defects, and in the case of unreasonable parameter setting, it may also fall into a local optimal solution. SUMMARY

[0007] In view of the problems in the prior art, the application provides a ship LAAV motion model parameter identification method based on a dynamic immune genetic algorithm (DIGA) with adaptive adjustment of hyperparameters. The ship LAAV motion model researched in the application aims to establish a space-time parameter model, that is, to simulate the real motion of a ship based on the trajectory mode of the ship, the environment information and the like, and to generate a model of actual motion data of the ship.

[0008] The application is implemented in the following manner: a ship LAAV motion model parameter identification method based on a dynamic immune genetic algorithm with adaptive adjustment of hyperparameters comprises the following steps.

[0009] Step 1: using actual sea trial data of a ship, constructing an identification model fitness criterion function, designing a hyperparameter adaptive adjustment mechanism, and determining model parameters through selection, crossover, mutation, vaccination and elite reservation of an initial population for several generations.

[0010] Step 2: verifying by experiment that the DIGA method is used to optimize unknown parameters in the ship LAAV motion model based on actual measurement data of the ship, so that the unknown parameters converge to an optimal solution, and the optimal solution is brought into the LAAV model to simulate and generate AVP parameters.

[0011] Further, the LAAV model comprises the following.

[0012] An LAAV mathematical function expression.

[0013] Ship AVP parameter calculation.

[0014] Further, the LAAV mathematical function expression comprises the following.

[0015] (1) a linear acceleration expression;

[0016] (2) an angular velocity model.

[0017] Further, the linear acceleration expression comprises the following.

[0018] Longitudinal linear acceleration a x

[0019] When sailing at a constant speed, the longitudinal linear acceleration is zero, that is, a x = 0.

[0020] When the ship is accelerating, the ship's speed is affected by the propeller thrust and hydrodynamic force. According to the a&ω model, when the ship type is constant, the mathematical expression of the longitudinal speed of the ship can be approximated as a second-order system time domain function, i.e. the mathematical expression of the longitudinal speed during acceleration is:

[0021]

[0022] In the formula, v0 is the initial speed of the ship during acceleration, ΔV is the longitudinal speed change of the ship during acceleration, T1 and T2 are time constants, and their expressions are as follows:

[0023]

[0024] In the formula, ω1 is the natural frequency, ξ1 is the damping ratio, and is an unknown parameter to be identified; the expression of the known longitudinal linear speed v x is known, and the longitudinal linear acceleration expression is obtained by derivation:

[0025]

[0026] When the ship is decelerating, the longitudinal linear speed change can be regarded as the inverse process of the acceleration process.

[0027] The lateral linear acceleration a y

[0028] When the ship is straight sailing, the lateral speed change is not considered, and the lateral linear acceleration is zero, i.e. a y = 0.

[0029] When the ship is rotating, the longitudinal linear speed is regarded as a constant, i.e. only the change of the lateral linear acceleration during the constant rotation process is considered; according to the a&ω model, the lateral linear acceleration of the ship during the constant rotation is:

[0030] a y = v x ω z (4)

[0031] The vertical linear acceleration a z

[0032] When the ship is sailing on the water surface, the change of the vertical movement of the ship is very small, i.e. the vertical linear speed v z is zero, and therefore the vertical linear acceleration can be represented as a z = 0.

[0033] Further, the angular velocity model is:

[0034] The heading angular velocity ω z

[0035] The heading angle of the ship has little change during straight sailing, and thus the heading angle velocity ω z = 0.

[0036] During steady turning of the ship, the heading angle velocity can be expressed by a second-order time-domain function, which can be:

[0037]

[0038] In the formula, K is an unsigned proportional constant, δ is a rudder angle, T3 and T4 are time constants, K, ξ3 and ω3 are unknown parameters to be identified.

[0039] The roll angle velocity ω x

[0040] When the ship sails straight in still water, the roll angle velocity can be considered as zero, that is, ω x = 0.

[0041] When the ship turns, the change of the roll angle with time can be expressed by a second-order under-damped function:

[0042]

[0043] In the formula, γ max is the maximum roll angle during turning, β1, ξ4 and ω4 are the initial phase, damping ratio and natural frequency of the under-damped system respectively, and β1 = arccosξ4; ξ4 and ω4 are unknown parameters to be identified.

[0044] After the turning of the ship is completed, the roll angle gradually decreases, and its change can be considered as a cosine exponential decay function, and thus can be expressed as:

[0045] γ(t) = γ0e -k·t cos(ω5t+β2) (7)

[0046] In the formula, γ0 is the roll angle at the end of turning, k is an exponential decay coefficient, ω5 and β2 are the cosine exponential decay frequency and initial phase respectively, and β2 = arccosξ5; k, ξ5 and ω5 are unknown parameters to be identified.

[0047] According to the above analysis, the roll angle velocity can be obtained by derivation of the formula (6) and (7), and thus the expression of the roll angle velocity is:

[0048]

[0049] The pitch angle velocity ω y

[0050] When the ship is in uniform straight sailing motion, the pitch angle is approximately zero, so the pitch angle velocity is ω y = 0.

[0051] When the ship is in variable speed straight sailing motion, the pitch angle can be expressed by a second-order over-damped function due to the action of the propeller and the rudder:

[0052]

[0053] wherein θ0 is the initial pitch angle of the ship, Δθ is the pitch angle change of the ship in the variable speed process, ξ6, ω6 and β3 are the cosine exponential decay damping ratio, the cosine exponential decay frequency and the initial phase, respectively, wherein β3 = arccosξ6; ξ6, ω6 are unknown parameters to be identified.

[0054] Further, the ship AVP parameter solving:

[0055] The ship's attitude A(k), speed V(k) and position P(k) at time k are respectively:

[0056]

[0057] In formula (10), is the roll angle at time k, θ(k) is the pitch angle at time k, ψ(k) is the heading angle at time k; v X (k) is the eastward speed at time k, v Y (k) is the northward speed at time k, v Z (k) is the skyward speed at time k; λ(k) is the longitude at time k, L(k) is the latitude at time k, H(k) is the height at time k.

[0058] Attitude update

[0059] The ship's attitude A(k+1) at time k+1 is:

[0060]

[0061] In formula (11), ω x (k) is the roll angle velocity of the ship at time k, ω y (k) is the pitch angle velocity of the ship at time k, ω z (k) is the heading angle velocity of the ship at time k; ts is the update step; is the coefficient matrix of the Euler angle differential equation,

[0062]

[0063] In formula (12) represents the roll angle, θ represents the pitch angle, and ψ represents the yaw angle.

[0064] Velocity update

[0065] The carrier velocity V of the ship at time k+1 b (k+1) is expressed as:

[0066]

[0067] In order to facilitate the calculation of position information, the b system velocity needs to be converted into the n system velocity V n (k+1):

[0068] Wherein is a direction cosine matrix

[0069]

[0070] Position update

[0071] The position P(k+1) of the ship at time k+1 is:

[0072]

[0073] Taking the average velocity as the position update velocity, In formula (52), M can be expressed as:

[0074]

[0075] R M and R N are the radii of curvature of the meridian and the prime vertical of the point where the ship is located, respectively;

[0076]

[0077] Another object of the present application is to provide a ship LAAV motion model parameter identification system based on an immune genetic algorithm, comprising:

[0078] A model parameter determination module is configured to use actual sea trial data of the ship to construct an identification model fitness criterion function, and to determine model parameters by performing selection, crossover, mutation, vaccination and elite reservation on an initial population for several generations.

[0079] An AVP parameter generation module is configured to use the DIGA method to optimize unknown parameters in the ship LAAV motion model with respect to actual measurement data of the ship, so that the unknown parameters converge to an optimal solution, and the optimal solution is brought into the LAAV model to simulate and generate AVP parameters.

[0080] Another object of the present application is to provide a computer device comprising a memory and a processor, the memory storing a computer program, the computer program being executed by the processor to cause the processor to perform the steps of the ship LAAV motion model parameter identification method based on a dynamic immune genetic algorithm.

[0081] Another object of the present application is to provide a computer readable storage medium storing a computer program, the computer program being executed by a processor to cause the processor to perform the steps of the ship LAAV motion model parameter identification method based on a dynamic immune genetic algorithm.

[0082] Another object of the present application is to provide an information data processing terminal comprising the ship LAAV motion model parameter identification system based on a dynamic immune genetic algorithm.

[0083] In combination with the above technical solutions and the technical problems solved, the technical solutions to be protected by the present application have the following advantages and positive effects:

[0084] First, the ship LAAV motion model researched by the present application is different from other ship maneuvering (control) models, and aims to establish a space-time parameter model, i.e., to simulate the real motion of a ship based on the planned route of the ship, the trajectory mode, the environmental information, etc., and to generate a model of the actual motion parameters of the ship.

[0085] The present application aims to realize a ship motion data generator that is closer to the characteristics of a real ship, and therefore, with reference to a six-degree-of-freedom ship model, a six-degree-of-freedom LAAV motion model of a ship is proposed, the linear acceleration and angular velocity variation of the ship are described by mathematical expressions, and then the AVP parameters of the ship are obtained by algorithm solving.

[0086] The application aims at the problems existing in the parameter identification of ship LAAV motion model: the single initial value setting of the static identification method is easy to cause the overflow of the optimization result, resulting in low identification accuracy; in addition, only a single set of data can be input for static identification each time, and there are defects such as low identification efficiency and long time consumption. An innovative identification framework based on dynamic immune genetic algorithm (DIGA) is provided. Through the fusion of the super parameter cooperative optimization mechanism, the performance bottleneck of the traditional static algorithm is broken through. The DIGA framework proposes an outer optimization and inner identification working mechanism. The outer layer is used for optimizing the super parameter combination such as the crossover rate and the mutation rate, and minimizing the identification error of the inner layer; at the same time, the population diversity in the inner layer algorithm drives the evolution of the super parameter of the outer layer, and the inner layer executes the dynamic identification of the model parameters under the given super parameter. It is found through test comparison that the identification accuracy of DIGA is increased by 83.37% and 26.21% than SIGA and DGA respectively, and the convergence speed is increased by 75% and 60% than SIGA and static IGA respectively. The AVP data generated by the LAAV model after algorithm solving is equivalent to the performance of the a&omega model generator, and the similarity with the test data is more than 90%.

[0087] In order to solve the identification problem of multiple parameters to be determined in the ship motion data model, a method for realizing adaptive identification of unknown parameters in the LAAV model by DIGA is provided. Through the DIGA intelligent algorithm, the ship measured data is used, the ship maneuvering state information is taken as the input data, the fitness function is constructed, the super parameter adaptive adjustment mechanism is designed, and the operations such as cross, mutation, vaccine extraction and vaccination are combined, the elite is reserved and the population is updated, and the individual with the highest fitness is found to realize the identification purpose.

[0088] In order to verify the feasibility of the scheme, the measured data not participating in the identification is selected to compare and verify the data generated by the LAAV model, and it can be seen from the comparison result that the data generated by the trajectory generator of the LAAV model identified by DIGA has small deviation from the measured data, the change curves of the trajectory, the speed and the heading angle are relatively smooth, and are consistent with the actual change situation.

[0089] The modeling method for identifying unknown parameters in the mathematical expression of the ship LAAV model and generating AVP parameters by the solving method provided by the application has the following technical advantages:

[0090] 1) The DIGA framework has obvious advantages compared with the SIGA and DGA methods, and has the characteristics of fast convergence speed and high identification accuracy in processing the nonlinear system with multiple unknown parameters to be identified.

[0091] 2) The generation of AVP data can reflect the actual motion situation of the ship. The data generated by the established LAAV model can generate the motion data of the ship under the static water condition, and provides a data source and a benchmark for navigation research.

[0092] 3) Refer to the trajectory generator design scheme of the application, the parameter identification can be carried out on the measured data of different types of ships, and the LAAV motion model of a specific ship type can be established. In this way, the LAAV models of different types of ships can be established, and the generated ship motion data can be used for simulation work in the related navigation and ship field, which has practical significance.

[0093] Secondly, the traditional ship motion data generation technology has the following limitations: precision defects: the trajectory generator based on empirical formula (such as PROFGEN) has a fitting degree of less than 70% with the actual motion data; the identification accuracy of the static parameter identification method (such as static immune genetic algorithm (SIGA)) is relatively large (MSE>9%).

[0094] Filling way: first dynamic immune genetic framework: 6-DOF parameter identification error is less than 1.8% (traditional method>9%).

[0095] Patent novelty basis (compared with US20210073421A1, CN113609417A): for the first time, the super parameter adaptive mechanism is introduced into the ship motion model parameter identification; the LAAV model architecture under static water mode is realized.

[0096] The "double-high dilemma" of ship motion model parameter identification: the contradiction between high precision requirement (RMSE<2%) and high-dimensional nonlinear system solving efficiency. Traditional method: intelligent algorithm (standard GA / PSO) is easy to fall into local optimum (occurrence rate>40%); numerical optimization method (such as SQP) has exponential growth of computational complexity.

[0097] Breakthrough solution: super parameter adaptive adjustment: linear adjustment strategy is adopted to optimize the crossover rate (Pc:0.85→0.34) and mutation rate (Pm:0.08→0.2) in real time; vaccine enhancement mechanism: through the gene fragment memory (MemoryBank) of the elite population, the global search efficiency is improved by 3.2 times; hierarchical identification strategy: the 6-DOF parameters are decoupled into linear acceleration / angular velocity independent subspaces, reducing the problem dimension.

[0098] Practical verification: in the real ship data test, the test comparison found that the identification accuracy of DIGA was increased by 83.37% and 26.21% compared with SIGA and DGA respectively, and the convergence speed was increased by 75% and 60% compared with SIGA and static IGA respectively; the AVP data generated by the LAAV model after algorithm calculation can reach 98% similarity with the data generated by the a&ω model generator, and the similarity with the test data is more than 90%. BRIEF DESCRIPTION OF DRAWINGS

[0099] Figure 1A flow chart of a ship LAAV motion model parameter identification method based on a dynamic immune genetic algorithm (DIGA) is provided in the embodiments of the present application.

[0100] Figure 2 A system structure block diagram of a ship LAAV motion model parameter identification system based on a dynamic immune genetic algorithm (DIGA) is provided in the embodiments of the present application.

[0101] Figure 3 An LAAV model diagram is provided in the embodiments of the present application.

[0102] Figure 4 A ship coordinate system diagram is provided in the embodiments of the present application.

[0103] Figure 5 A DIGA parameter identification flow chart is provided in the embodiments of the present application.

[0104] Figure 6 An identification result of two unknown parameters in an acceleration mathematical model is provided in the embodiments of the present application; a comparison diagram of the identification results of the parameters in an acceleration section under three methods of DIGA, SIGA and DGA; (a) a comparison diagram of MSE results, (b) a comparison diagram of fitting curves.

[0105] Figure 7 An identification result of three unknown parameters in a rotation mathematical model is provided in the embodiments of the present application; a comparison diagram of the identification results of the heading angle velocity under 30° rudder angle under three methods of DIGA, SIGA and DGA; (a) a comparison diagram of MSE in the identification process, (b) a comparison diagram of fitting curves.

[0106] Figure 8 An AVP parameter simulation diagram is provided in the embodiments of the present application; (a) a heading angle, (b) a trajectory, (c) a northeast sky velocity, (d) a track speed.

[0107] Figure 9 A comparison diagram of ship sea trial data and simulation AVP data is provided in the embodiments of the present application; (a) a trajectory comparison diagram, (b) a heading angle comparison diagram, (c) an eastward velocity comparison diagram, (d) a northward velocity comparison diagram; (e) a track speed comparison diagram.

[0108] Figure 10 An effect diagram of a linear self-adaptive adjustment strategy of hyperparameters (a crossover rate, a mutation rate) is provided in the embodiments of the present application. DETAILED DESCRIPTION

[0109] In order to make the objectives, technical solutions and advantages of the present application clearer, the present application is further described in detail below with reference to embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application, and are not used to limit the present application.

[0110] Traditional ship LAAV kinematic model generally relies on fixed architecture genetic algorithm or least square method in parameter identification stage. The former is prone to premature convergence in high-dimensional coupled parameter space, and the latter is numerically unstable in the presence of measurement noise and model nonlinearity. Especially under the triple coupling effect of fast variable load of propeller, large disturbance of rudder angle and irregular wave field, the longitudinal, lateral and yaw angular velocity show significant time-varying characteristics. Static algorithm is difficult to capture the real-time drift of dynamic coefficients synchronously, which directly leads to the accumulation of attitude-velocity-position (AVP) solving error and the increase of path reconstruction deviation.

[0111] To solve the above bottleneck, this scheme introduces dynamic immune genetic algorithm (DIGA) in the basic framework of gene crossover and mutation, and embeds antibody concentration inhibition and affinity acceleration mechanism. The algorithm adaptively adjusts the crossover probability, mutation probability and immune threshold according to the population diversity in each generation, forming a "convergence-diffusion" breathing rhythm: when the population aggregation is too high, the mutation strength is improved to avoid local extremum; when the global exploration tends to be sparse, the crossover and immune affinity are improved to compress the search domain and maintain the global optimal attraction.

[0112] When the method is landed, first use multi-source sea trial data such as accelerometer, gyroscope, velocity against ground radar to construct multi-objective fitness criteria. The residual square of longitudinal, lateral and vertical linear acceleration, phase difference of yaw angular velocity, peak leakage of roll angular velocity, etc. are weighted and summed to form the objective function, and the weight is dynamically reconstructed according to the working condition of the ship. The higher the fitness, the smaller the fitting error and the better the dynamic consistency. The population individuals are selected by affinity selection and antibody memory unit screening, ensuring that the high fitness genes are preserved.

[0113] In each iteration step of population evolution, the unknown parameters of the longitudinal velocity second-order system model, the lateral roll under-damped model and the yaw second-order response model are injected into the LAAV equation cluster in real time. After the equation is discretized by the fourth-order Runge-Kutta method, a short window AVP sequence is generated and compared with the measured sequence. If the residual rises rapidly, the algorithm triggers immune mutation operation to slightly disturb the high affinity individuals; if the residual decreases slowly, it enters the fine tuning stage, and only applies elite crossover to low affinity individuals to strengthen the local search density.

[0114] When the affinity amplitude is lower than the set threshold and there is no significant improvement for several generations, it is considered to be converged. At this time, the natural frequency, damping ratio, proportional gain and other parameters obtained are fixed and written into the LAAV model, and then the complete AVP curve is generated in the full time domain. The simulation results are aligned with the sea trial data, and the root mean square error of the three-axis attitude is significantly converged to the millimeter level, and the velocity-position coupling error is less than 3%, which meets the requirements of steering and path reconstruction accuracy.

[0115] Compared with the existing static genetic algorithm and particle swarm method, the DIGA significantly improves the global search efficiency through the immune memory-inhibition-mutation triple mechanism; the fitness-driven super parameter adjustment breaks the fixed probability constraint and avoids premature convergence; and the dynamic affinity screening ensures the fast tracking ability of the algorithm under complex wave-rudder-propeller combined interference. The final result not only improves the applicability of the LAAV model to extreme sea conditions, but also reduces the cost of repeated sea trial of the ship, and provides a high-reliability dynamic foundation for ship motion control and fault diagnosis.

[0116] As Figure 1 shown, the ship LAAV motion model parameter identification method based on the dynamic immune genetic algorithm provided by the embodiment of the application comprises the following steps:

[0117] S101, using the actual sea trial data of the ship, constructing an identification model fitness criterion function, designing a super parameter adaptive adjustment mechanism, and determining the model parameters through selection, crossover, mutation, vaccination and elite reservation of the initial population for several generations;

[0118] S102, the DIGA method is used to optimize the unknown parameters in the ship LAAV motion model for the measured data of the ship, so that it converges to the optimal solution, and the optimal solution is brought into the LAAV model to generate the AVP parameters.

[0119] The LAAV model provided by the embodiment of the application comprises:

[0120] LAAV mathematical function expression;

[0121] Ship AVP parameter solution.

[0122] The LAAV mathematical function expression provided by the embodiment of the application comprises:

[0123] (1) linear acceleration expression;

[0124] (2) angular velocity model.

[0125] The linear acceleration expression provided by the embodiment of the application comprises:

[0126] Longitudinal linear acceleration a x

[0127] When sailing at a constant speed, the longitudinal linear acceleration is zero, that is, a x =0.

[0128] When the ship is accelerating, the speed of the ship is affected by the propeller thrust and the hydrodynamic force. According to the a&omega model, when the ship type is constant, the mathematical expression of the longitudinal speed of the ship can be approximated as a two-order system time domain function, that is, the mathematical expression of the longitudinal speed during acceleration is:

[0129]

[0130] In the formula, v0 is the initial speed in the acceleration process of the ship, ΔV is the longitudinal speed change in the acceleration process of the ship, T1 and T2 are time constants, and the expressions are as follows:

[0131]

[0132] In the formula, ω1 is the natural frequency, ξ1 is the damping ratio, the longitudinal linear speed v is known, and the expression is obtained by derivation of the longitudinal linear acceleration expression: x

[0133]

[0134] When the ship is decelerating, the longitudinal linear speed change can be regarded as the inverse process of the acceleration process.

[0135] The transverse linear acceleration a y

[0136] When the ship is in straight sailing, the transverse linear acceleration is zero, that is, a y = 0.

[0137] When the ship is in rotation, the longitudinal linear speed is regarded as a constant, that is, only the change of the transverse linear acceleration in the process of the steady rotation is considered; according to the a&ω model, the transverse linear acceleration of the ship in the steady rotation is:

[0138] a y = v x ω z (22)

[0139] The vertical linear acceleration a z

[0140] When the ship sails on the water surface, the change of the vertical movement of the ship is very small, that is, the vertical linear speed v z is zero, therefore, the vertical linear acceleration can be represented as a z = 0.

[0141] The angular velocity model provided by the embodiment of the application is:

[0142] The heading angular velocity ω z

[0143] The heading angle of the ship hardly changes in the straight sailing process, so the heading angular velocity is ω z = 0.

[0144] In the steady rotation process of the ship, the heading angular velocity can be represented by a second-order system time domain function, which can be:

[0145] ​

[0146] where K is a non-dimensional proportionality constant, δ is the rudder angle, T3 and T4 are time constants, ω3 is the natural frequency of the turning, and ξ3 is the damping ratio of the turning.

[0147] Rolling angular velocity ω x

[0148] When the ship is sailing straight in still water, the rolling angular velocity can be considered as zero, i.e. ω x = 0.

[0149] When the ship is turning, the change of the rolling angle with time can be expressed by a second-order under-damped function:

[0150]

[0151] where γ max is the maximum roll angle during the turning, β1, ξ4 and ω4 are the initial phase, damping ratio and natural frequency of the under-damped system, respectively, and β1 = arccosξ4.

[0152] After the turning of the ship is finished, the rolling angle will gradually decrease, and its change can be considered as a cosine exponential decay function, which is expressed as:

[0153] γ(t) = γ0e -k·t cos(ω5t+β2) (25)

[0154] where γ0 is the rolling angle at the end of the turning, k is the exponential decay coefficient, ω5 and β2 are the cosine exponential decay frequency and initial phase, respectively, and β2 = arccosξ5, ξ5 is the damping ratio in the cosine exponential decay function.

[0155] According to the above analysis, the rolling angular velocity can be obtained by differentiating the equations (24) and (25), and thus the expression of the rolling angular velocity is:

[0156]

[0157] Pitching angular velocity ω y

[0158] When the ship is sailing straight at a constant speed, the pitching angle is almost unchanged and can be considered as zero, and thus the pitching angular velocity is ω y = 0.

[0159] When the ship is sailing straight at a variable speed, the moments acting on the ship due to the propeller and the rudder will cause the ship to pitch, and the pitching angle can be expressed by a second-order over-damped function:

[0160]

[0161] wherein θ0 is an initial pitch angle of the ship, Δθ is a pitch angle change of the ship during the speed change process, ξ6, ω6 and β3 are a cosine exponential decay damping ratio, a cosine exponential decay frequency and an initial phase, respectively, wherein β3 = arccosξ6.

[0162] The ship AVP parameter solving method provided by the embodiment of the present application comprises the following steps:

[0163] The ship attitude A(k), speed V(k) and position P(k) at time k are respectively:

[0164]

[0165] Attitude updating

[0166] The ship attitude at time k+1 is A(k+1):

[0167]

[0168] wherein ts is an updating step;

[0169] is a coefficient matrix of the Euler angle differential equation,

[0170]

[0171] wherein represents a roll angle, θ represents a pitch angle, and ψ represents a yaw angle;

[0172] Speed updating

[0173] The ship carrier speed V b (k+1) at time k+1 is:

[0174]

[0175] In order to facilitate the solving of the position information, the b system speed needs to be converted into the n system speed V n (k+1):

[0176]

[0177] wherein is:

[0178]

[0179] Position updating

[0180] The ship position P(k+1) at time k+1 is:

[0181]

[0182] The average speed is used as the speed of position updating, In formula (52), M can be represented as:

[0183]

[0184] R M and R N are the radii of curvature of the meridian and prime vertical of the point where the ship is located respectively;

[0185]

[0186] As Figure 2 shown, the ship LAAV motion model parameter identification system based on a dynamic immune genetic algorithm (DIGA) provided by the embodiment of the application comprises:

[0187] A model parameter determination module is configured to use actual sea trial data of the ship, construct an identification model fitness criterion function, design a hyperparameter adaptive adjustment mechanism, and determine model parameters through selection, crossover, mutation, vaccination and elite reservation of an initial population for several generations.

[0188] An AVP parameter generation module is configured to use the DIGA method to optimize unknown parameters in the ship LAAV motion model for the actual measured data of the ship, so that the unknown parameters converge to an optimal solution, and the optimal solution is brought into the LAAV model to simulate and generate AVP parameters.

[0189] The computer device provided by the embodiment of the application comprises a memory and a processor, and the memory stores a computer program.

[0190] The computer readable storage medium provided by the embodiment of the application stores a computer program.

[0191] The information data processing terminal provided by the embodiment of the application is used to implement the ship LAAV motion model parameter identification system based on the dynamic immune genetic algorithm.

[0192] The embodiment of the application is specifically implemented as follows:

[0193] 1. Ship LAAV motion model

[0194] The LAAV model comprises a motion mathematical function expression and AVP parameter solving; unknown parameters in the mathematical function expression are solved by using the preprocessed ship measured sailing data through the DIGA identification algorithm; and the unknown parameters after solving are brought into the LAAV mathematical function expression to generate the AVP parameters through the algorithm. The model structure is shown in Figure 3

[0195] The ship LAAV model is a mathematical model used to represent the motion changes of the linear acceleration and angular acceleration of the ship in the track coordinate system. The coordinate system definition is shown in Figure 4

[0196] According to the above coordinate system definition, the six-degree-of-freedom linear acceleration and angular velocity of the ship are defined as follows: x is the longitudinal linear acceleration, y is the transverse linear acceleration, z is the vertical linear acceleration; ω x is the roll angular acceleration, y is the pitch angular acceleration, and z is the heading angular acceleration.

[0197] 1.1 LAAV mathematical function expression

[0198] The mathematical expressions of the linear acceleration and angular velocity of the ship in different motion states will be given below. The LAAV mathematical function expression comprises a linear acceleration function expression and an angular velocity function expression. For convenience of introduction, the derivation process of the LAAV mathematical function model will not be described in detail.

[0199] 1.1.1 Linear acceleration expression

[0200] The longitudinal linear acceleration a x

[0201] When the ship sails at a constant speed, the longitudinal linear acceleration is zero, that is, a x = 0.

[0202] When the ship sails at an acceleration, the speed of the ship is affected by the propeller thrust and the hydrodynamic force. According to the a&ω model, when the ship type is constant, the mathematical expression of the longitudinal speed of the ship can be approximated as a two-order system time domain function, that is, the mathematical expression of the longitudinal speed of the ship when accelerating is:

[0203] ​​

[0204] where v0 is the initial speed of the ship during the acceleration process, ΔV is the longitudinal speed variation of the ship during the acceleration process, and T1 and T2 are time constants, and their expressions are as follows:

[0205]

[0206] where ω1 is the natural frequency and ξ1 is the damping ratio. Given the expression of the longitudinal linear speed v x , the expression of the longitudinal linear acceleration is obtained by derivation:

[0207]

[0208] When the ship is in deceleration, the longitudinal linear speed variation can be regarded as the inverse process of the acceleration process.

[0209] The lateral linear acceleration a y

[0210] When the ship is in straight sailing, the lateral linear acceleration is zero, i.e., a y = 0.

[0211] When the ship is in rotation, the longitudinal linear speed is regarded as a constant, i.e., only the variation of the lateral linear acceleration in the process of steady rotation is considered. According to the a&ω model, the lateral linear acceleration of the ship in steady rotation is:

[0212] a y = v x ω z (40)

[0213] The vertical linear acceleration a z

[0214] When the ship is sailing on the water surface, the variation of the ship in the vertical direction is very small, i.e., the vertical linear speed v z is zero, and thus the vertical linear acceleration can be expressed as a z = 0.

[0215] 1.1.2 Angular velocity model

[0216] The heading angular velocity ω z

[0217] During the straight sailing of the ship, the heading angle almost does not change, and thus the heading angular velocity is ω z = 0.

[0218] During the steady rotation of the ship, the heading angular velocity can be expressed by a second-order system time domain function, which can be:

[0219]

[0220] where K is a non-negative proportional constant, δ is the rudder angle, T3 and T4 are time constants,

[0221] Rolling angle velocity ω x

[0222] When the ship is sailing straight in still water, the rolling angle velocity can be considered as zero, i.e. ω x = 0.

[0223] When the ship is making a turning motion, the change of the rolling angle with time can be expressed by a second order under-damped function:

[0224]

[0225] where γ max is the maximum roll angle during the turning process, β1, ξ4 and ω4 are the initial phase, damping ratio and natural frequency of the under-damped system respectively, where β1 = arccos ξ4.

[0226] After the turning motion of the ship is finished, the rolling angle will gradually decrease, and its change can be considered as a cosine exponential decay function, so it can be expressed as:

[0227] γ(t) = γ0e -k·t cos(ω5t + β2) (43)

[0228] where γ0 is the rolling angle at the end of the turning, k is the exponential decay coefficient, ω5 and β2 are the cosine exponential decay frequency and initial phase respectively, where β2 = arccos ξ5.

[0229] According to the above analysis, the rolling angle velocity can be obtained by differentiating the equations (11) and (12), so the expression of the rolling angle velocity is:

[0230]

[0231] Pitching angle velocity ω y

[0232] When the ship is sailing straight at a constant speed, the pitching angle is almost unchanged and can be considered as zero, so the pitching angle velocity is ω y = 0.

[0233] When the ship is sailing straight at a variable speed, the moments acting on the ship due to the propeller and the rudder will cause the ship to pitch, and the pitching angle can be expressed by a second order over-damped function:

[0234]

[0235] where θ0is the initial pitch angle of the ship, Δθ is the change of the pitch angle of the ship during the speed change process, ξ6, ω6and β3are the cosine exponential decay damping ratio, the cosine exponential decay frequency and the initial phase respectively, and β3= arccosξ6.

[0236] 1.2 Ship AVP parameter solving

[0237] The attitude, velocity and position of the ship at time k are respectively:

[0238]

[0239] Attitude update

[0240] The attitude of the ship at time k+1 is:

[0241]

[0242] where ts is the update step.

[0243] is the coefficient matrix of the Euler angle differential equation,

[0244]

[0245] where denotes the roll angle, θ denotes the pitch angle, and ψ denotes the yaw angle.

[0246] Velocity update

[0247] The velocity of the ship at time k+1 is represented as:

[0248]

[0249] In order to facilitate the solving of position information, the b-based velocity needs to be converted into n-based velocity:

[0250]

[0251] where is:

[0252]

[0253] Position update

[0254] The position of the ship at time k+1 is:

[0255]

[0256] The average velocity is taken as the velocity for position update, In equation (52), M can be represented as:

[0257]

[0258] R M and R N respectively are the radii of curvature of the meridian and prime vertical at the point where the ship is located;

[0259]

[0260] 2 Dynamic identification framework of immune genetic algorithm

[0261] 2.1 Description of parameters to be identified

[0262] The parameters to be identified in the mathematical model of the ship LAAV are: the damping ratio ξ1 and the natural frequency ω1 in formulas (1) and (2) during acceleration; the proportional coefficient K, the damping ratio ξ3 and the natural frequency ω3 in formulas (5) and (6) during rotation. The deceleration process is regarded as the inverse process of the acceleration process, and its unknown parameters do not have to be identified separately. According to the preliminary analysis of the measured data, the change of the roll angle and the change of the pitch angle of a large ship sailing on a calm water surface are small, and in the present application, the parameter identification work is not discussed.

[0263] Therefore, the parameters to be identified in the present application are five, namely ξ1, ω1, K, ξ3 and ω3, and through the exploration of their physical meanings, the initial range of the parameters to be identified is shown in Table 1.

[0264] Table 1 Initial range of parameters to be identified

[0265]

[0266]

[0267] 2.2 Key technologies of dynamic identification of immune genetic algorithm

[0268] The static parameter identification method, such as the standard GA / IGA, faces two major bottlenecks in the identification of such a complex nonlinear system as the LAAV motion model: one is the super parameter fixation: the super parameters such as the crossover rate and the mutation rate need to be artificially preset and cannot adapt to the dynamic identification process; the other is the premature convergence: the single optimization level leads to the decline of the population diversity and falls into the local optimum. Therefore, the goal of the proposed dynamic immune genetic algorithm (DIGA) framework is to construct an adaptive dynamic identification framework to realize the collaborative optimization of the super parameters and the model parameters.

[0269] The dynamic immune genetic algorithm proposed in the present application mainly has two key technical strategies: one is the super parameter adaptive adjustment strategy, and the other is the immune mechanism enhancement strategy.

[0270] The implementation idea of the super parameter adaptive adjustment strategy is to dynamically adjust the super parameter crossover rate p of the genetic algorithm by monitoring the diversity change of the population in real time.m and the mutation rate p c , which is negatively correlated with the population diversity, so as to balance the global search and local exploitation ability of the algorithm. When the population diversity decreases, the mutation probability is increased to enhance the global search ability; when the diversity is high, the mutation probability is appropriately reduced to accelerate convergence; the adjustment of the crossover probability is opposite.

[0271] The vaccine enhancement mechanism converts historical search experience or prior knowledge into gene fragments by introducing the concept of "vaccine" in the biological immune system, actively guiding the evolution direction of the population, thereby accelerating convergence and improving recognition accuracy.

[0272] The core idea of the vaccine enhancement mechanism is: first, according to the biological inspiration: simulate the preventive effect of vaccines in the immune system, inject part of the parameters of known excellent individuals into the population to avoid blind search; second, dynamic learning: the vaccine is not fixed, but is dynamically extracted from the optimal solution accumulated during the algorithm running; and third, balance strategy: vaccination, mutation and crossover operation are coordinated to avoid over-reliance on vaccines leading to premature convergence.

[0273] 2.3 Dynamic identification framework implementation process

[0274] The dynamic immune genetic algorithm (DIGA) dynamically balances global exploration and local development according to population diversity, and can achieve efficient parameter search and identification of nonlinear systems in the LAAV mathematical model through adaptive adjustment of the crossover rate, mutation rate and immune mechanism (vaccination, antibody inhibition). By giving an initial population, after a limited number of selection, crossover and mutation, the gene fragments (vaccine) of the historical optimal solution are introduced to accelerate convergence. The final population can quickly converge, and the elements in the individual with the smallest MSE are the identified values of the parameters in the model.

[0275] To visually display the identification results, the present application uses real number coding, and the parameters to be identified are regarded as gene fragments in the individual. The five parameters to be identified in the present application require real-time data from two different data sources, namely the track speed and the heading angle speed. Therefore, two sets of dynamic identification work are required during the identification process. The first set is to process the track speed data to solve the unknown parameters xi and omega1 in equation (1), so the population individual is defined as xi = [xi, omega1]. The second set is to process the heading angle data to solve the unknown parameters K, xi and omega1 in equation (5), so the individual can be defined as x2 = [K, xi, omega3]. The population size is N = 50, and to ensure accurate optimization of the algorithm, the maximum number of iterations M = 200.

[0276] The specific steps of identifying the unknown parameters in the LAAV mathematical model using the dynamic immune genetic algorithm (DIGA) are as follows.

[0277] (1) Population initialization: according to the initial range interval [X min ,X max ] of the to-be-identified parameters in Table 1, an initial population pop0 is set by using a random interpolation method, so that the population is randomly distributed in the range interval, and the initial value of the population is:

[0278] X i =X min +(X max -X min )×k ( 55)

[0279] In the formula, i is the population sequence number (i = 1, 2,..., N), and k is a random number between 0 and 1.

[0280] (2) Set the fitness criterion function: find a set of parameters to make the time series of the calculated data and the actual data have the highest similarity to achieve the purpose of parameter identification, that is, the error between the simulation data Y e and the measured data Y is minimized. In the parameter identification problem, the mean squared error (MSE) is a statistical index for measuring the difference between the model prediction value and the actual observation value, so the MSE is usually used as the fitness function (Fitness Function) to drive the algorithm to search for the optimal parameters. The smaller the MSE of each candidate parameter set, the higher the fitness.

[0281] MSE j =mean(∑(Y j -Y ej ) 2 ) ( 56 )

[0282] (3) Adaptive hyperparameter adjustment.

[0283] This part combines two adjustment mechanisms to dynamically adjust the crossover rate and mutation rate. Mechanism one is linear adjustment based on the number of generations, which gradually reduces the crossover rate and increases the mutation rate with the number of iterations. Mechanism two is adjustment based on population diversity, which further suppresses crossover and enhances mutation when the population diversity is insufficient. According to the evolution generation (gen) and the population diversity (diversity), the parameters are adjusted adaptively to meet the following rules: early iteration: high crossover rate, low mutation rate (fast exploration of solution space). Late iteration: low crossover rate, high mutation rate (fine search, avoid premature convergence).

[0284] ① Hyperparameter initialization. To meet the early iteration rule: high initial crossover rate to encourage gene exchange, and low initial mutation rate to avoid destroying good genes. Set the reference parameters as: initial crossover rate p base,c = 0.85, and initial mutation rate pbase,m = 0.08.

[0285] 2. Linear adjustment based on algebra.

[0286]

[0287] 3. Adjustment based on population diversity. If the population diversity (the degree of dispersion of gene distribution) is lower than the threshold value (10% of the parameter range), it indicates that the population may fall into a local optimum, and the following strategies need to be taken:

[0288] Reduce the crossover rate: reduce ineffective gene exchange and focus on optimizing the current area.

[0289] Increase the mutation rate: increase randomness and explore new solution areas.

[0290] 4. Boundary protection. To prevent extreme parameter values from causing algorithm instability, the hyperparameters need to be stabilized within a certain boundary threshold. In the research method of the present invention, the crossover rate p c = [0.34, 0.85] and the mutation rate p m = [0.08, 0.2].

[0291] (4) Genetic operation.

[0292] 1. Selection: Since this identification problem needs to maintain diversity to avoid local optimum, roulette wheel selection is considered to select parent individuals P1, P2 with higher fitness to reproduce offspring, while ensuring population diversity.

[0293] 2. Crossover: The arithmetic crossover method is used, and the specific operation expression of the two parents is as follows:

[0294]

[0295] where α is a random weight used to generate offspring between the two parents. When α < p c , the crossover operation is performed according to the above formula.

[0296] 3. Mutation: Mutation operation is responsible for introducing new genetic diversity, helping the algorithm to jump out of the local optimal solution and increasing the possibility of finding the global optimal solution. When the random weight is less than the mutation rate, Gaussian mutation operation is performed on the population. The operation expression of the Gaussian mutation process is as follows:

[0297] pop new = pop old + σ·rand() (59)

[0298] where σ = 0.1*(X max -X min ), used to control the disturbance amplitude to ensure that the mutation is within a reasonable range; X is the parameter to be identified, Xmax is the upper bound of the range of the to-be-identified parameter, X min is the lower bound of the range; rand() is a random number between 0 and 1.

[0299] (5) Immunization operation. The purpose is to inject the gene fragments of the historical optimal individual into the current population with a certain probability, guide the search direction, and accelerate the convergence.

[0300] ① Extraction of vaccine: based on the fitness analysis of the initial population, the individuals with the top 30% fitness values are listed in the elite population, and part of the gene fragments (unknown parameter latitude to be identified) of the individuals are selected to generate a vaccine.

[0301] ② Vaccination: according to the vaccination rules, select the appropriate vaccination probability, and perform the vaccination operation on the individuals in the population. Replace the gene fragments at the corresponding positions in the individual chromosomes with the extracted vaccine fragments to obtain the vaccinated population pop v .

[0302] (6) Elite preservation mechanism (update population): a certain number of the best individuals in each generation are preserved to prevent the loss of good genes during evolution. When the population is updated, the first half of the population directly preserves the elite individuals of the current generation, and the second half of the population is selected from the newly generated offspring to obtain the remaining number of individuals.

[0303] (8) Termination condition: determine whether the termination condition is met, i.e., whether the maximum number of iterations is reached. If the termination condition is met, output the current optimal solution; if not, return to step (3) and continue the next iteration.

[0304] The DIGA parameter identification flowchart is shown in Figure 5 .

[0305] 3 Test verification

[0306] 3.1 IGA identification of unknown parameters

[0307] To verify the identification accuracy and convergence efficiency of the dynamic immune genetic algorithm, in view of the measured sea trial data of a certain ship, about four hundred thousand measured data points covering various acceleration and deceleration and turning sections and other different maneuvering states are selected, and the static immune genetic algorithm (SIGA), the dynamic genetic algorithm (DGA) and the DIGA method proposed in the present application are used to identify the five unknown parameters in the LAAV mathematical model.

[0308] In the experiment, the speed variation mathematical model and the heading angle speed variation mathematical model are two different nonlinear functions, so in the process of identification, they are taken as two independent identification mechanisms to carry out identification work. First, take the speed model identification as an example, in order to ensure the effectiveness of comparison, in addition to introducing the super parameter adaptive adjustment operation, the rest of the algorithm design is the same, the difference between SGA and DIGA is only whether there is the participation of immune mechanism, and the super parameter adaptive adjustment mechanism is strictly the same. The comparison experiment parameter setting is shown in Table 2.

[0309] Table 2 Comparison experiment setting

[0310]

[0311] After setting the initial conditions, the three kinds of algorithms will carry out 15 times of independent identification work on the same set of original data (eliminate the influence of randomness), record the convergence curve, precision and calculation time of each run in the identification process, and then carry out statistical analysis, and the following comparison charts are obtained after analysis.

[0312] The following is the identification result comparison of the mean value taken after identification by the three kinds of algorithms on a section of acceleration segment.

[0313] Figure 6 The parameter identification result for the acceleration segment.

[0314] In addition, the unknown parameter identification result in the heading angle speed model under the condition of 30° full rudder is as follows.

[0315] Figure 7 The parameter identification result for the heading angle speed.

[0316] The statistical analysis result is directly shown by Table 3, and through the data in the table, it can be found that the convergence speed of DIGA algorithm is the fastest and the identification precision is the highest, which can show that the algorithm is accurate and reliable in LAAV mathematical model identification.

[0317] Table 3 Comparison of parameter identification results of three kinds of algorithms

[0318]

[0319] According to the identification results of several measured motion data, the ship's track speed change is divided into five modes, and five track speed thresholds (unit: kn) 0.3, 1, 5, 10, 16 are specified. The threshold value of the speed change and the corresponding unknown parameter are given in Table 4. The speed change below 0.3kn can be regarded as uniform speed, and the remaining change between the two thresholds, the first half of the parameter is consistent with the previous threshold parameter, and the second half is consistent with the previous threshold parameter. For example, the speed change between 0.3kn and 1kn, the parameter value of the first half (0.3kn-0.65kn) is taken as the value at 0.3kn, and the parameter value of the second half is taken as the value at 1kn, and the parameter values of ξ1 and ω1 are selected in the same way.

[0320] Table 4 Correspondence table of speed change threshold and unknown parameter of a certain ship

[0321]

[0322] During the turning process, different rudder angles will bring different turning radii. According to the actual navigation situation, the turning process is divided into three modes, and the rudder angle thresholds are 10°, 20° and 30°. The rudder angle threshold and its corresponding unknown parameter are shown in Table 5. From the actual operation of the ship, the rudder angle below 10° belongs to the small rudder angle range, which is used for fine adjustment of the heading during the ship's navigation, and a long distance is needed to change the heading. Therefore, the rudder angle below 10° can be regarded as 0 degree, which does not change the heading angle. The full rudder value of the ship is 30°, and the LAAV model specifies that the unknown parameter value in mode 1 is taken for the rudder angle of 10°-15°, the unknown parameter value in mode 2 is taken for the rudder angle of 15°-25°, and the unknown parameter value in mode 3 is taken for the rudder angle of 25°-30°.

[0323] Table 5 Identification results of rudder angle threshold parameters

[0324]

[0325] 3.2 AVP data simulation

[0326] According to the identification results in section 3.1, the LAAV motion model of a certain ship can be obtained. Only the initial position, attitude and speed information of the ship need to be determined, and the preset ship maneuvering state is input, and the actual motion of the ship can be simulated by the LAAV motion model.

[0327] Let the initial attitude angle and speed of the ship be 0, and the initial position be east longitude 120°, north latitude 15°. In order to fully reflect the various maneuvering states of the ship, the total navigation time of the ship is set to 1 hour. The navigation scheme is designed as follows:

[0328] (1) The speed change is 3kn, the rudder angle is 0, and the navigation time is 400 seconds;

[0329] (2) The speed change is 6 knots, the rudder angle is 0, and the sailing time is 600 seconds;

[0330] (3) The speed change is -4kn, the rudder angle is 0, and the sailing time is 300 seconds;

[0331] (4) The speed change is -1 knot, the rudder angle is 26°, and the sailing time is 400 seconds;

[0332] (5) The speed change is 16 knots, the rudder angle is 0, and the sailing time is 200 seconds;

[0333] (6) The change in speed is 0, the rudder angle is 0, and the sailing time is 400 seconds;

[0334] (7) The speed change is -0.3kn, the rudder angle is -14°, and the sailing time is 300 seconds;

[0335] (8) The speed change is -7kn, the rudder angle is 0, and the sailing time is 500 seconds;

[0336] (9) The change in speed is 0, the rudder angle is 0, and the sailing time is 200 seconds;

[0337] (10) The speed change is 3 knots, the rudder angle is 0, and the sailing time is 200 seconds.

[0338] The simulation results of the AVP parameters of this scheme are as follows (the red circle in Figure (b) indicates the trajectory starting point):

[0339] Depend on Figure 8 As can be seen, based on the input maneuvering state, the LAAV model can simulate and generate the AVP motion parameters under the expected conditions: heading angle, longitude, latitude, and northeast-southeast velocity.

[0340] 3.3 Feasibility Verification of the LAAV Model

[0341] To verify the feasibility of the LAAV model, this invention compares the AVP parameters generated by the LAAV model with simulation data from the a&ω model data generator. The model data generator overcomes the shortcomings of traditional trajectory generators, which produce curves with significant deviations from measured data; its accuracy in position, speed, pitch, roll, and heading all exceed 90%. Therefore, the data generated by the LAAV model can be compared with it to verify the model's feasibility.

[0342] To ensure the rigor of the verification test, the LAAV model verification part selects the measured data not used for DIGA identification to participate in the verification. The identification results in Tables 2 and 3 are respectively brought into the LAAV mathematical model of the ship, and the initial position of the ship in the data segment for verification is 122.8656 ° E, 30.8818 ° N, the longitudinal velocity is 9.1082 kn, the lateral velocity and the skyward velocity are both 0, and the heading angle is -92.42 °. The specific maneuvering state is shown in Table 6.

[0343] Table 6 Ship maneuvering state

[0344]

[0345] The above maneuvering state is input into the LAAV model to generate the corresponding AVP motion parameters, which are compared with the a&ω model generated data. The comparison results are shown in Figure 9

[0346] Figure 9 Ship sea trial data and simulation AVP data comparison chart

[0347] Through the comparison of the five charts in Figure 9 , it can be seen that the unknown parameter values obtained by the DIGA identification framework are brought into the LAAV model, and the simulation generated AVP data has little difference with the a&ω model generated data and the measured data, and the curve fitting degree is high.

[0348] After a large number of measured and simulated data comparison and analysis, it is found that the AVP parameters generated by the model have the following data generation conditions: trajectory up to 91%, heading angle data fitting degree 98%, and track speed curve fitting degree 93%. It can be seen that the DIGA identification result is accurate, and the LAAV model has good engineering practical application ability and can generate AVP parameters that fit the actual ship motion.

[0349] 4 Conclusion

[0350] To solve the identification problem of multiple parameters to be determined in the ship motion data model, a method of using DIGA to realize adaptive identification of unknown parameters in LAAV model is proposed. Through the DIGA intelligent algorithm, the ship maneuvering state information is used as input data, the fitness function is constructed, the super parameter adaptive adjustment mechanism is designed, and the operations such as cross, mutation, vaccine extraction and vaccination are combined to update the population constantly, find the individual with the highest fitness to achieve the identification purpose.

[0351] ​To verify the feasibility of the scheme, the measured data not involved in identification is selected to compare with the data generated by the LAAV model, and the comparison result can be seen that the deviation between the data generated by the LAAV model trajectory generator built by DIGA identification and the measured data is small, the change curves of the trajectory, the sailing speed and the heading angle are relatively smooth, and the actual change situation is consistent.

[0352] The modeling method for generating AVP parameters by combining the solving method with the unknown parameters in the mathematical expression of the ship LAAV model based on the DIGA identification has the following technical advantages:

[0353] 1) Compared with the SIGA and DGA methods, the DIGA framework has obvious advantages, and has the characteristics of fast convergence speed and high identification accuracy in processing the nonlinear system with multiple unknown parameters to be identified. The algorithm framework design can be widely used in the optimization and identification demand of the nonlinear system, and the real-time identification function can be considered in the future.

[0354] 2) The generation of AVP data can reflect the actual motion of the ship. The comparison and verification of the simulation results show that the design scheme of the DIGA identification LAAV model parameter algorithm is feasible, and the data generated by the established LAAV model can generate the motion data of the ship under the condition of still water, and provide data source and benchmark for navigation research.

[0355] 3) According to the design scheme of the trajectory generator, the measured data of different types of ships can be identified, and the LAAV motion model of a specific ship type can be established. In this way, the LAAV models of different types of ships can be established, and the generated ship motion data can be used for simulation work in the related navigation and ship field, which has practical significance.

[0356] The present application is based on the research under the condition of still water, and in the actual situation, the ship navigation will be disturbed by environmental factors, and the ship motion model under the influence of marine factors should be considered in the subsequent research work.

[0357] I. The specific application field of the present application or related products.

[0358] 1. Navigation algorithm development and test tool chain

[0359] Application scenario: can replace the traditional sea trial, and provide a low-cost test environment for path following (Path Following), model predictive control (MPC) and other algorithms.

[0360] Related products: NavSimPro simulation platform: cloud SaaS service, provides API interface to generate customized AVP data set (such as uncontrolled ship motion under typhoon weather); algorithm performance evaluation suite: quantitatively analyzes the trajectory tracking error (RMSE) of the navigation algorithm.

[0361] 2. Intelligent navigation and autonomous navigation system

[0362] Application scenario: Real-time motion modeling and prediction capabilities can be provided for unmanned cargo ships and intelligent tugboats to support dynamic collision avoidance decisions; high-precision port approach and departure trajectory data can be generated to empower port AGV (automatic guided vehicle) collaborative scheduling systems.

[0363] Related products: Ship digital twin engine: an embedded system integrating DIGA-LAAV model, real-time output of 6-DOF motion data stream, hardware adaptation edge computing platform; autonomous navigation decision module: reinforcement learning controller based on AVP data, supporting path re-planning in dynamic environment.

[0364] 3. Maritime training and safety assessment

[0365] Application scenario: Crew simulator upgrade: generate extreme condition training data to improve the authenticity of emergency operation training; navigation safety prediction: reconstruct collision / grounding scenarios based on historical accident data to assess the defects of ship maneuvering systems.

[0366] Related products: Intelligent navigation simulator: VR / AR compatible training system supporting multi-ship cooperative exercise; safety risk assessment software: generate 10^5 level accident scenario library through Monte Carlo simulation, calculate risk probability distribution (output SIL safety level).

[0367] 4. Ocean engineering and special ships

[0368] Application scenario: Engineering ship operation simulation: simulate complex operations such as semi-submersible ship heavy diving, pipe-laying ship dynamic positioning (DP), etc.; naval ship tactical test: generate concealed maneuvering trajectories (such as Z-type evasion) to support shipborne weapon system fire control algorithm optimization.

[0369] Related products: Engineering ship digital operation manual: an interactive system based on LAAV model, real-time prediction of lifting operation stability boundary; military ship tactical simulator: generate multi-dimensional motion data in confrontation scenarios (including electromagnetic silence / electronic jamming mode).

[0370] II. Related evidence of the technical effects obtained by the embodiments of the present application.

[0371] 1. Evidence of algorithm performance advantage

[0372] Test design:

[0373] Take the speed model identification as an example, in order to ensure the effectiveness of comparison, in addition to introducing the super parameter adaptive adjustment operation, the rest of the algorithm design is the same between SIGA and DIGA, the difference between SGA and DIGA is only whether the immune mechanism is involved or not, and the super parameter adaptive adjustment mechanism is strictly consistent. The comparison experiment parameter setting is shown in Table 7.

[0374] Table 7 Comparison test setting

[0375]

[0376] After setting the initial conditions, the three kinds of algorithms will perform 15 times of independent identification work on the same set of original data (eliminate the influence of randomness), record the convergence curve, accuracy and calculation time of each run during the identification process, and then perform statistical analysis.

[0377] Core data:

[0378] Figure 6 Acceleration segment parameter identification result;

[0379] Figure 7 Heading angle velocity parameter identification result;

[0380] The statistical analysis results are directly shown by Table 8.

[0381] Table 8 Comparison of parameter identification results of three kinds of algorithms

[0382]

[0383] Conclusion support:

[0384] The test comparison found that the identification accuracy of DIGA was increased by 83.37% and 26.21% for SIGA and DGA respectively, and the convergence speed was increased by 75% and 60% for SIGA and static IGA respectively.

[0385] Through the data in the table, it can be found that the convergence speed of DIGA algorithm is the fastest and the identification accuracy is the highest, which can show that the algorithm is accurate and reliable in LAAV mathematical model identification.

[0386] 2、Engineering effectiveness verification evidence

[0387] Test design:

[0388] In order to verify the feasibility of the LAAV model, the AVP parameters generated by the LAAV model are compared with the simulation data of the a&omega model data generator. The model data generator breaks through the defect that the generated curve of the traditional trajectory generator has obvious deviation from the measured data, and the accuracy of position, speed, pitch, roll and heading is more than 95%. Therefore, the generated data of the LAAV model can be compared with it to verify the feasibility of the model.

[0389] To ensure the rigor of the verification experiment, the LAAV model verification section selected measured data that were not used for DIGA identification. The identification results in Tables 2 and 3 were substituted into the ship's LAAV mathematical model. In the verification data segment, the ship's initial position was 122.8656°E, 30.8818°N, the longitudinal speed was 9.1082kn, the lateral speed and azimuth speed were both 0, and the heading angle was -92.42°. The specific maneuvering states are shown in Table 9.

[0390] Table 9 Ship Maneuver Status

[0391]

[0392] The above maneuvering conditions are input into the LAAV model to generate the corresponding AVP motion parameters, which are then compared with the data generated by the a&ω model.

[0393] Key data:

[0394] The comparison results are as follows Figure 9 As shown, through Figure 9 The five comparison charts clearly show that, after the unknown parameter values ​​obtained through the DIGA identification framework are substituted into the LAAV model, the simulation-generated AVP data is not significantly different from the data generated by the a&ω model and the measured data, and the curve fit is high.

[0395] Through extensive comparative analysis of measured and simulated data, the AVP parameters generated by this model show the following data generation characteristics: trajectory accuracy reaches 91%, heading angle data fit is 98%, and track speed curve fit is 93%. This demonstrates that the DIGA identification results are accurate, and the LAAV model possesses excellent practical engineering application capabilities, generating AVP parameters that closely match actual ship motion.

[0396] The AVP data generated by the LAAV model after algorithmic calculation has a similarity of up to 98% with the data generated by the a&ω model generator, and a similarity of over 90% with the experimental data.

[0397] Supporting conclusions:

[0398] The LAAV model is highly feasible, and its simulation-generated data can be used in navigation algorithm research and other related application fields.

[0399] It should be noted that embodiments of the present application can be realized by hardware, software, or a combination of software and hardware. The hardware portion can be realized by a special logic; the software portion can be stored in a memory and executed by a proper instruction execution system, such as a microprocessor or a specially designed hardware. A person of ordinary skill in the art can understand that the above-mentioned apparatus and method can be realized by computer executable instructions and / or included in processor control codes, such as a carrier medium, such as a magnetic disk, CD or DVD-ROM, a programmable memory, such as a read-only memory (firmware), or a data carrier, such as an optical or electronic signal carrier. The apparatus of the present application and its modules can be realized by a hardware circuit, such as a very large scale integrated circuit or a gate array, a semiconductor, such as a logic chip, a transistor, or a programmable hardware device, such as a field programmable gate array, a programmable logic device, or the like, by software executed by various types of processors, or by a combination of the above-mentioned hardware circuit and software, such as firmware.

[0400] The above description is merely a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any modification, equivalent replacement, and improvement within the technical range disclosed by the present application, and within the spirit and principle of the present application, should be covered within the protection scope of the present application.

Claims

1. A ship LAAV motion model parameter identification method based on a dynamic immune genetic algorithm, characterized in that, The method comprises the following steps: Collecting ship measured sea trial data and establishing a fitness criterion function; Setting up a fitness-driven hyperparameter self-adaptive adjustment mechanism, after initializing the population, sequentially executing selection, crossover, mutation, immunity and elite retention to obtain the optimal parameters of the model; Substituting the obtained optimal parameters into the LAAV motion model to simulate and generate AVP parameters for ship attitude, speed and position calculation.

2. The method according to claim 1, characterized in that, The LAAV motion model comprises a linear acceleration sub-model and an angular velocity sub-model, the linear acceleration sub-model describes longitudinal, lateral and vertical linear accelerations, and the angular velocity sub-model describes heading angular velocity, roll angular velocity and pitch angular velocity.

3. The method according to claim 2, characterized in that, The longitudinal linear acceleration is modeled as a second-order time-domain function based on the coupling effect of ship propeller thrust and hydrodynamic force, the function parameters are determined by natural frequency and damping ratio, and the longitudinal acceleration expression is obtained by derivation.

4. The method according to claim 2, characterized in that, The heading angular velocity is modeled as a second-order time-domain function of proportional coefficient and rudder angle drive in steady turning state, the roll angular velocity is modeled as an under-damped system function during turning and a cosine exponential decay function after turning.

5. The method according to claim 1, characterized in that, The crossover probability, mutation probability and immune threshold of the dynamic immune genetic algorithm are self-adaptively adjusted according to the current population fitness distribution, so as to improve the global optimization ability and convergence speed.

6. A ship LAAV motion model parameter identification system based on a dynamic immune genetic algorithm, characterized in that, It comprises: A parameter determination module for executing steps one and two of claim 1 to obtain the optimal parameters of the model; An AVP parameter generation module for inputting the optimal parameters into the LAAV model and outputting the AVP parameters.

7. A system according to claim 6, characterised in that It also comprises a data acquisition interface for receiving ship sea trial sensor data in real time and providing it to the parameter determination module.

8. The system according to claim 6, characterized in that The parameter determination module sequentially comprises a fitness evaluation unit, a selection unit, a crossover unit, a mutation unit, an immunity unit and an elite retention unit, and the control parameters of each unit are updated in real time through the hyperparameter self-adaptive adjustment mechanism.

9. A storage medium having instructions stored thereon, the instructions being executed on a processor to implement the method of claim 1.

10. The storage medium according to claim 9, characterized in that The instructions also cause the processor to complete ship attitude update, speed update and position update in the Euler integral manner based on the AVP parameters.

Images