An unmanned ship control system integrating multiple communication links and interfaces

Abstract

The present invention relates to the field of unmanned ship control technology, and discloses an unmanned ship control system integrating multiple communication links and interfaces, including: a data acquisition module, used to collect environmental parameters, control parameters and state parameters of the unmanned ship during navigation, and represent them through a first feature sequence; a preprocessing and feature extraction module, used to process the first feature sequence to obtain a second feature sequence; an unmanned ship state prediction module, used to predict the position, roll angle, pitch angle and power of the unmanned ship in a second preset time period according to a preset unmanned ship state prediction model; a control parameter optimization module, used to optimize the control parameters of the unmanned ship according to an improved genetic algorithm to obtain the optimal control parameters. The present invention realizes dynamic optimization of control parameters such as speed, heading angle and power distribution, and can adapt to complex and changing environments in real time.

Description

Technical Field

[0001] The present invention belongs to the technical field of unmanned ship control, and in particular relates to an unmanned ship control system integrating multiple communication links and interfaces. Background Art

[0002] With the widespread application of unmanned ships in hydrological monitoring, environmental protection, military reconnaissance, logistics and transportation, their autonomous control technology has become the focus of related research. The control system of unmanned ships usually needs to achieve precise navigation, stable navigation and efficient energy management in complex and changeable water environments. However, due to various environmental factors (such as wind speed, wave height, water flow direction) and equipment characteristics (such as propeller power), the autonomous navigation of unmanned ships faces many challenges.

[0003] In the existing technology, the optimization of control parameters of unmanned ships (such as speed, heading angle and power distribution) is the key to improving system performance. However, traditional control methods mostly rely on static models and fail to fully consider the real-time changes of dynamic environmental factors, resulting in insufficient adaptability of unmanned ships in complex environments. For example, when the water flow velocity and wind speed change, failure to adjust the control parameters in real time will cause the unmanned ship to deviate from the route or navigate unstably. In addition, existing optimization methods often ignore the coupling relationship between control parameters and the comprehensive optimization of the stability and energy consumption of unmanned ships, further limiting the performance improvement of unmanned ships. Summary of the invention

[0004] The present invention provides an unmanned ship control system integrating multiple communication links and interfaces, which solves the technical problem in related technologies that the unmanned ship cannot adapt to dynamic environmental changes in real time, resulting in limited adaptability and performance of the unmanned ship in complex environments.

[0005] The present invention provides an unmanned ship control system integrating multiple communication links and interfaces, including:

[0006] A data acquisition module, used for collecting environmental parameters, control parameters and state parameters of the unmanned ship during navigation within a first preset time period at a first preset time interval, and representing them through a first characteristic sequence;

[0007] Environmental parameters include: water velocity, water direction, wave height, wave propagation direction, wind speed and wind direction;

[0008] Control parameters include: speed, heading angle, power distribution;

[0009] State parameters include: position, roll angle, pitch angle and power;

[0010] A preprocessing and feature extraction module, used for processing the first feature sequence to obtain a second feature sequence, wherein the nth sequence unit of the second feature sequence represents the nth preprocessed environmental parameter, control parameter and state parameter;

[0011] An unmanned ship state prediction module, used to predict the position, roll angle, pitch angle and power of the unmanned ship in a second preset time period according to a preset unmanned ship state prediction model, the input of the unmanned ship state prediction model is a second feature sequence, and the output is the position, roll angle, pitch angle and power of the unmanned ship in the second preset time period;

[0012] The control parameter optimization module is used to optimize the control parameters of the unmanned ship according to the improved genetic algorithm to obtain the optimal control parameters.

[0013] Furthermore, sine and cosine functions are used to double encode the water flow direction, wind direction and heading angle, the roll angle and pitch angle are represented by real number encoding, the power distribution is represented by the power ratio of the left and right thrusters, and the position is represented by latitude and longitude coordinates.

[0014] Further, the first feature sequence is processed to obtain a second feature sequence, and the specific steps include:

[0015] Step 201: extract wind flow disturbance synthesis features according to the first feature sequence, and concatenate them after the first feature sequence. The calculation formula of wind flow disturbance synthesis features is:

[0016] ,in, represents the synthetic characteristics of wind flow disturbance, Indicates the water flow velocity, Indicates the direction of water flow. Indicates wind speed, Indicates wind direction. represents the angle between the water flow and the wind direction, and t represents the first preset time interval;

[0017] Step 202: extract the wind-wave coupling power consumption feature according to the first feature sequence, and concatenate it after the first feature sequence. The calculation formula of the wind-wave coupling power consumption feature is:

[0018] ,in, represents the wind-wave coupling power consumption characteristics, represents the density of water, Indicates wave height, It represents the angle between wind direction and wave height propagation direction. It represents the angle between the water flow direction and the wave height propagation direction. Indicates the difference between the wave height at the current time point and the wave height at the previous time point;

[0019] Step 203, extract the comprehensive feature of speed and heading angle according to the first feature sequence, and splice it after the first feature sequence. The calculation formula of the comprehensive feature of speed and heading angle is:

[0020] ,in, It represents the comprehensive characteristics of speed and heading angle. Indicates the speed. represents the heading angle, Indicates the difference between the speed at the current time point and the speed at the previous time point;

[0021] Step 204: extract the dynamic stability feature of the hull according to the first feature sequence and concatenate it after the first feature sequence. The calculation formula of the dynamic stability feature of the hull is:

[0022] ,in, represents the dynamic stability characteristics of the hull, N represents the number of time points, a, b, c and d represent the indexes of the time points, and denote the roll angles at the ath time point and the bth time point, respectively. and denote the pitch angles at the cth time point and the dth time point, respectively. Indicates the difference between the roll angle at the current time point and the roll angle at the previous time point. Indicates the difference between the pitch angle at the current time point and the pitch angle at the previous time point. represents the pitch angle at the ath time point;

[0023] Step 205: Use the z-score normalization method to normalize the concatenated first feature sequence to obtain a second feature sequence.

[0024] Furthermore, the Pareto principle is used to improve the long short-term memory network, and an unmanned ship state prediction model is constructed based on the improved long short-term memory network. The unmanned ship state prediction model includes: an input feature screening layer, a grouping feature processing layer, a feature fusion layer and an output layer;

[0025] The input feature screening layer is used to divide the features of the sequence units in the second feature sequence into key features and secondary features according to the Pareto principle, and obtain a key feature sequence and a secondary feature sequence respectively;

[0026] The grouping feature processing layer is used to process the key feature sequence and the secondary feature sequence respectively;

[0027] The feature fusion layer is used to splice the processing results of the key feature sequence and the secondary feature sequence to obtain the global feature;

[0028] The output layer is used to input the global features into the first classifier, the second classifier, the third classifier and the fourth classifier respectively, and the classification spaces of the first classifier, the second classifier, the third classifier and the fourth classifier respectively represent the position, the roll angle, the pitch angle and the power.

[0029] Further, according to the Pareto principle, the features of the sequence units in the second feature sequence are divided into key features and secondary features, and a key feature sequence and a secondary feature sequence are obtained respectively. The specific steps include:

[0030] Step 301, calculating the Pearson correlation coefficient score between each feature of the sequence unit in the second feature sequence and the state parameter;

[0031] Step 302, arranging the calculation results of step 301 in descending order;

[0032] Step 303, selecting the first K features sorted in step 302 as key features, and the remaining features as secondary features, and finally obtaining a key feature sequence and a secondary feature sequence.

[0033] Furthermore, the calculation formula of the Pearson correlation coefficient score in step 301 is:

[0034] ;

[0035] in, represents the Pearson correlation coefficient score of the e-th feature of the second feature sequence, The number of features representing the state parameters, represents the index of the feature of the state parameter, N represents the number of time points, f, g and h represent the time point index, and represent the feature values ​​of the e-th feature of the second feature sequence at the f-th time point and the g-th time point, respectively. and They represent the preset values ​​of the j-th feature of the state parameter at the f-th time point and the h-th time point respectively, where the preset values ​​represent predicted feature values ​​and are calculated based on a linear equation constructed using historical data.

[0036] Further, the grouping feature processing layer is used to process the key feature sequence and the secondary feature sequence respectively. Specifically, the key feature sequence is processed using the first network layer to obtain a first hidden state, and the secondary feature sequence is processed using the second network layer to obtain a second hidden state;

[0037] Among them, the calculation formula of the sth sequence unit of the key feature sequence of the first network layer includes:

[0038] ;

[0039] ;

[0040] ;

[0041] ;

[0042] ;

[0043] ;

[0044] in, , , , , and They represent the input gate, forget gate, output gate, candidate memory unit, memory unit update and hidden state update of the sth sequence unit of the key feature sequence respectively. represents the sth sequence unit of the key feature sequence, and They represent the hidden states output by the s-th sequence unit and the s-1-th sequence unit, respectively. , and Respectively represent the first weight parameter, the first bias parameter and the second weight parameter of the input gate, represents the enhanced weight of the memory cell state of the previous sequence unit, Indicates that the sth sequence unit and the hidden state output by the s-1th sequence unit To splice, represents the memory cell update of the s-1th sequence unit, , and Respectively represent the first weight parameter, the first bias parameter and the second weight parameter of the forget gate, , and Respectively represent the first weight parameter, the first bias parameter and the second weight parameter of the output gate, and They represent the first weight parameter and the first bias parameter of the candidate memory unit respectively, sigmoid represents the sigmoid activation function, and tanh represents the hyperbolic tangent activation function;

[0045] Extract the hidden state output by the Nth sequence unit as the first hidden state;

[0046] The second network layer removes the input gate and output gate, candidate memory cells, and retains the forget gate, memory cell update, and hidden state update.

[0047] Furthermore, the control parameters of the unmanned ship are optimized according to the improved genetic algorithm, and the specific steps include:

[0048] Step 401, generating an initialization population including P individuals that meets the constraint conditions based on a spatial hierarchical method;

[0049] Step 402, constructing an objective function according to the unmanned ship state prediction model;

[0050] Step 403, calculating the fitness value of each individual in the population according to the objective function;

[0051] Step 404, retaining V individuals with the highest fitness in the current population, and randomly selecting two individuals from them as parents;

[0052] Step 405, generating a random number ranging from 0 to 1, and when the random number is less than a first preset threshold, performing a crossover operation that satisfies the constraint condition, wherein the crossover operation represents selecting an element value of one position of the first feature vector corresponding to two individuals in the parent generation to be exchanged, wherein the first threshold is a custom parameter;

[0053] Step 406, updating the population individuals based on the differential mutation strategy;

[0054] Step 407, determine whether the current number of iterations is greater than or equal to the maximum number of iterations, and use the individual with the largest fitness value in the initialized population as the optimal control parameter, otherwise return to step 403, where the maximum number of iterations is a custom parameter.

[0055] Furthermore, the calculation formula of the objective function is:

[0056] ;

[0057] in, represents the fitness value of the objective function, , and represent the weight coefficients of the control distance target, control stability target and control energy consumption respectively, and , and The sum is 1, and They respectively represent the longitude of the current individual's unmanned ship position obtained by the unmanned ship state prediction model and the longitude of the individual's target position. and They respectively represent the latitude of the current individual's unmanned ship position obtained by the unmanned ship state prediction model and the latitude of the individual's target position. Represents a positive value between 0 and 1. Indicates the preset slope coefficient, used to control The slope of and They represent the roll angle and pitch angle of the unmanned ship obtained by the current individual through the unmanned ship state prediction model, represents the weight coefficient of the pitch angle relative to the roll angle, It represents the power of the unmanned boat obtained by the current individual through the unmanned boat state prediction model.

[0058] Furthermore, the specific steps of step 406 include:

[0059] Step 501, randomly select Y individuals from the population as individuals to be mutated;

[0060] Step 502, randomly selecting three different benchmark individuals from the population;

[0061] Step 503, updating the mutated individuals using the first updating formula according to the three selected benchmark individuals, the first updating formula is: ,in, represents the characteristic value of the individual with mutation, , and They represent the eigenvalues ​​of the three reference individuals, including the position, roll angle, pitch angle and power of the first eigenvector. Indicates the scaling factor.

[0062] The beneficial effects of the present invention are as follows: the present invention combines the three core objectives of distance, stability and energy consumption, and comprehensively optimizes the performance of the unmanned ship through flexible weight adjustment and unmanned ship state prediction model; through the improved genetic algorithm, it ensures rapid approach to the target point while improving navigation stability, and effectively reduces energy consumption, so as to adapt to the requirements of different complex tasks;

[0063] The present invention adopts a simplified long short-term memory network to process minor features and reduce unnecessary computational burden. At the same time, an improved long short-term memory network is used for key features to ensure high-precision modeling of important features, taking into account both accuracy and efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] Figure 1 It is a diagram of an unmanned ship control system integrating multiple communication links and interfaces of the present invention;

[0065] Figure 2 It is a flow chart of processing the first feature sequence according to the present invention. DETAILED DESCRIPTION

[0066] The subject matter described herein will now be discussed with reference to example embodiments. It should be understood that the discussion of these embodiments is only to enable those skilled in the art to better understand and implement the subject matter described herein, and the functions and arrangements of the elements discussed may be changed without departing from the scope of protection of the contents of this specification. Each example may omit, replace or add various processes or components as needed. In addition, the features described relative to some examples may also be combined in other examples.

[0067] like Figure 1-Figure 2 As shown, an unmanned ship control system integrating multiple communication links and interfaces includes:

[0068] The data collection module 101 is used to collect environmental parameters, control parameters and state parameters of the unmanned ship during navigation within a first preset time period at a first preset time interval, and represent them through a first feature sequence;

[0069] Environmental parameters include: water velocity, water direction, wave height, wave propagation direction, wind speed and wind direction;

[0070] Control parameters include: speed, heading angle, power distribution;

[0071] State parameters include: position, roll angle, pitch angle and power;

[0072] A preprocessing and feature extraction module 102 is used to process the first feature sequence to obtain a second feature sequence, wherein the nth sequence unit of the second feature sequence represents the nth preprocessed environmental parameter, control parameter and state parameter;

[0073] The unmanned ship state prediction module 103 is used to predict the position, roll angle, pitch angle and power of the unmanned ship in the second preset time period according to a preset unmanned ship state prediction model, the input of the unmanned ship state prediction model is the second feature sequence, and the output is the position, roll angle, pitch angle and power of the unmanned ship in the second preset time period;

[0074] The control parameter optimization module 104 is used to optimize the control parameters of the unmanned ship according to the improved genetic algorithm to obtain the optimal control parameters.

[0075] In one embodiment of the present invention, a Doppler anemometer is used to measure the water flow velocity and direction, a wave Doppler radar sensor is used to obtain the wave height and the wave height propagation direction, an anemometer and a wind vane are used to monitor the wind speed and direction, and the speed, heading angle and power distribution of the unmanned ship are obtained through the control center of the unmanned ship. The heading angle indicates the current movement direction of the unmanned ship, and the power distribution indicates the power ratio of the left and right thrusters of the unmanned ship. The current position of the unmanned ship, that is, the current longitude and latitude, is obtained through the Beidou navigation system on the unmanned ship, the roll angle and pitch angle of the unmanned ship are collected through the inertial measurement unit, and the power is obtained through the controller of the thruster.

[0076] In one embodiment of the present invention, sine and cosine functions are used to double encode the water flow direction, wind direction and heading angle. For example, when the heading angle is 90°, the heading angle is encoded as ,Right now ; The roll angle and pitch angle are represented by real number codes, and the power distribution is represented by the power ratio of the left and right thrusters. For example, if the power of the left thruster and the right thruster are 500W and 700W respectively, the power distribution is , the position is represented by the longitude and latitude coordinates, and the power is represented by the sum of the left thruster power and the right thruster power.

[0077] In one embodiment of the present invention, the first feature sequence is processed to obtain the second feature sequence, and the specific steps include:

[0078] Step 201: extract wind flow disturbance synthesis features according to the first feature sequence, and concatenate them after the first feature sequence. The calculation formula of wind flow disturbance synthesis features is:

[0079] ,in, represents the synthetic characteristics of wind flow disturbance, Indicates the water flow velocity, Indicates the direction of water flow. Indicates wind speed, Indicates wind direction. represents the angle between the water flow and the wind direction, and t represents the first preset time interval;

[0080] Step 202: extract the wind-wave coupling power consumption feature according to the first feature sequence, and concatenate it after the first feature sequence. The calculation formula of the wind-wave coupling power consumption feature is:

[0081] ,in, represents the wind-wave coupling power consumption characteristics, represents the density of water, Indicates wave height, It represents the angle between wind direction and wave height propagation direction. It represents the angle between the water flow direction and the wave height propagation direction. Indicates the difference between the wave height at the current time point and the wave height at the previous time point;

[0082] Step 203, extract the comprehensive feature of speed and heading angle according to the first feature sequence, and splice it after the first feature sequence. The calculation formula of the comprehensive feature of speed and heading angle is:

[0083] ,in, It represents the comprehensive characteristics of speed and heading angle. Indicates the speed. represents the heading angle, Indicates the difference between the speed at the current time point and the speed at the previous time point;

[0084] Step 204: extract the dynamic stability feature of the hull according to the first feature sequence and concatenate it after the first feature sequence. The calculation formula of the dynamic stability feature of the hull is:

[0085] ,in, represents the dynamic stability characteristics of the hull, N represents the number of time points, a, b, c and d represent the indexes of the time points, and denote the roll angles at the ath time point and the bth time point, respectively. and denote the pitch angles at the cth time point and the dth time point, respectively. Indicates the difference between the roll angle at the current time point and the roll angle at the previous time point. Indicates the difference between the pitch angle at the current time point and the pitch angle at the previous time point. represents the pitch angle at the ath time point;

[0086] Step 205: Use the z-score normalization method to normalize the concatenated first feature sequence to obtain a second feature sequence.

[0087] In one embodiment of the present invention, the long short-term memory network is improved using the Pareto principle, and an unmanned ship state prediction model is constructed based on the improved long short-term memory network. The unmanned ship state prediction model includes: an input feature screening layer, a grouping feature processing layer, a feature fusion layer and an output layer;

[0088] The input feature screening layer is used to divide the features of the sequence units in the second feature sequence into key features and secondary features according to the Pareto principle, and obtain a key feature sequence and a secondary feature sequence respectively;

[0089] The grouping feature processing layer is used to process the key feature sequence and the secondary feature sequence respectively;

[0090] The feature fusion layer is used to splice the processing results of the key feature sequence and the secondary feature sequence to obtain the global feature;

[0091] The output layer is used to input the global features into the first classifier, the second classifier, the third classifier and the fourth classifier respectively, and the classification spaces of the first classifier, the second classifier, the third classifier and the fourth classifier respectively represent the position, the roll angle, the pitch angle and the power.

[0092] In one embodiment of the present invention, the features of the sequence units in the second feature sequence are divided into key features and secondary features according to the Pareto principle, and a key feature sequence and a secondary feature sequence are obtained respectively. The specific steps include:

[0093] Step 301, calculating the Pearson correlation coefficient score between each feature of the sequence unit in the second feature sequence and the state parameter;

[0094] Step 302, arranging the calculation results of step 301 in descending order;

[0095] Step 303, selecting the first K features sorted in step 302 as key features, and the remaining features as secondary features, and finally obtaining a key feature sequence and a secondary feature sequence. Preferably, K is set to 4.

[0096] In one embodiment of the present invention, the calculation formula of the Pearson correlation coefficient score in step 301 is:

[0097] ;

[0098] in, represents the Pearson correlation coefficient score of the e-th feature of the second feature sequence, represents the number of features of the state parameter. In this embodiment, is 4, represents the index of the feature of the state parameter, N represents the number of time points, f, g and h represent the time point index, and represent the feature values ​​of the e-th feature of the second feature sequence at the f-th time point and the g-th time point, respectively. and They represent the preset values ​​of the j-th feature of the state parameter at the f-th time point and the h-th time point respectively, where the preset values ​​represent predicted feature values ​​and are calculated based on a linear equation constructed using historical data.

[0099] In one embodiment of the present invention, the Pareto principle is used to select K features that have the greatest impact on the state parameters output by the model from environmental parameters, control parameters and state parameters, and then process them separately to improve the efficiency and accuracy of the model.

[0100] In one embodiment of the present invention, the group feature processing layer is used to process the key feature sequence and the secondary feature sequence respectively. Specifically, the key feature sequence is processed using the first network layer to obtain a first hidden state, and the secondary feature sequence is processed using the second network layer to obtain a second hidden state;

[0101] Among them, the calculation formula of the sth sequence unit of the key feature sequence of the first network layer includes:

[0102] ;

[0103] ;

[0104] ;

[0105] ;

[0106] ;

[0107] ;

[0108] in, , , , , and They represent the input gate, forget gate, output gate, candidate memory unit, memory unit update and hidden state update of the sth sequence unit of the key feature sequence respectively. represents the sth sequence unit of the key feature sequence, and They represent the hidden states output by the s-th sequence unit and the s-1-th sequence unit, respectively. , and Respectively represent the first weight parameter, the first bias parameter and the second weight parameter of the input gate, The enhanced weight representing the memory cell state of the previous sequence unit is used to enhance the sensitivity of the current unit to long-term memory. Indicates that the sth sequence unit and the hidden state output by the s-1th sequence unit To splice, represents the memory cell update of the s-1th sequence unit, , and Respectively represent the first weight parameter, the first bias parameter and the second weight parameter of the forget gate, Used to emphasize the nonlinear interaction between the current input and the hidden state. , and Respectively represent the first weight parameter, the first bias parameter and the second weight parameter of the output gate, The input gate and memory cell state are combined to dynamically weight the output. and They represent the first weight parameter and the first bias parameter of the candidate memory unit respectively, sigmoid represents the sigmoid activation function, and tanh represents the hyperbolic tangent activation function.

[0109] Extract the hidden state output by the Nth sequence unit as the first hidden state;

[0110] The second network layer removes the input gate and output gate, candidate memory cells, and retains the forget gate, memory cell update, and hidden state update;

[0111] The calculation formula for the u-th sequence unit of the second network layer includes:

[0112] ;

[0113] ;

[0114] ;

[0115] in, and They represent the forget gate and memory unit update of the u-th sequence unit of the secondary feature sequence, represents the u-th sequence unit of the secondary feature sequence, and They represent the second hidden states of the u-th sequence unit and the u-1-th sequence unit outputs, respectively. represents the memory unit update of the u-1th sequence unit of the secondary feature sequence, and Respectively represent the first weight parameter and the first bias parameter of the forget gate, and Represents the first weight parameter and the first bias parameter updated by the memory unit.

[0116] In one embodiment of the present invention, by using an improved long short-term memory network to process key feature sequences, complex dependencies in time series can be deeply explored to ensure high-precision modeling of important features; at the same time, a simplified version of the long short-term memory network is used to process secondary feature sequences, which reduces computational complexity and avoids waste of resources, thereby improving computational efficiency while ensuring prediction performance.

[0117] In one embodiment of the present invention, the control parameters of the unmanned ship are optimized according to the improved genetic algorithm, and the specific steps include:

[0118] Step 401, generating an initialization population containing P individuals that meet the constraint conditions based on a spatial hierarchical method, specifically, dividing the range of speed, heading angle and power allocation into P equally spaced sub-intervals, randomly sampling a value from each sub-interval to form P individuals, and combining the P individuals into an initialization population, where P is a custom parameter;

[0119] Initialize the individual encoding of the population through the first eigenvector Indicates that Respectively represent the speed, heading angle, and power distribution;

[0120] The constraints include:

[0121] Speed ​​range constraint: The speed of the unmanned ship needs to be within the effective range;

[0122] Heading angle range constraint: The heading angle of the unmanned ship needs to be within the valid range;

[0123] Step 402, constructing an objective function according to the unmanned ship state prediction model;

[0124] Step 403, calculating the fitness value of each individual in the population according to the objective function;

[0125] Step 404, retaining V individuals with the highest fitness in the current population, and randomly selecting two individuals from them as parents, preferably, V is set to 10;

[0126] Step 405, generating a random number ranging from 0 to 1, when the random number is less than a first preset threshold, performing a crossover operation that satisfies the constraint condition, wherein the crossover operation means selecting the element value of one position of the first feature vector corresponding to two individuals in the parent generation to exchange, if the result after the exchange does not satisfy the constraint condition, re-performing the crossover operation until the constraint condition is satisfied, wherein the first threshold is a custom parameter, preferably, the first threshold is set to 0.7;

[0127] Step 406, updating the population individuals based on the differential mutation strategy;

[0128] Step 407, determine whether the current number of iterations is greater than or equal to the maximum number of iterations, and use the individual with the largest fitness value in the initialized population as the optimal control parameter, otherwise return to step 403, where the maximum number of iterations is a custom parameter, preferably, the maximum number of iterations is set to 200.

[0129] In one embodiment of the present invention, the calculation formula of the objective function is:

[0130] ;

[0131] in, represents the fitness value of the objective function, , and represent the weight coefficients of the control distance target, control stability target and control energy consumption respectively, and , and The sum is 1, and They respectively represent the longitude of the current individual's unmanned ship position obtained by the unmanned ship state prediction model and the longitude of the individual's target position. and They respectively represent the latitude of the current individual's unmanned ship position obtained by the unmanned ship state prediction model and the latitude of the individual's target position. Represents a positive value, usually , exp represents the exponential function, Indicates the preset slope coefficient, used to control The slope of the unmanned ship is adjusted by experiment to select the optimal slope coefficient for optimizing the stability of the unmanned ship. and They represent the roll angle and pitch angle of the unmanned ship obtained by the current individual through the unmanned ship state prediction model, represents the weight coefficient of the pitch angle relative to the roll angle, , indicating that the two have the same impact on stability. When It represents the power of the unmanned boat obtained by the current individual through the unmanned boat state prediction model.

[0132] In one embodiment of the present invention, the objective function combines the three objectives of distance, stability and energy consumption, and flexibly adjusts the priority of each objective through weights to adapt to different task requirements, thereby achieving all-round optimization of the performance of the unmanned ship and balancing the requirements of rapid approach to the target point, stability and low energy consumption.

[0133] In one embodiment of the present invention, the specific steps of step 406 include:

[0134] Step 501, randomly select Y individuals from the population as individuals to be mutated;

[0135] Step 502, randomly selecting three different benchmark individuals from the population;

[0136] Step 503, updating the mutated individuals using the first updating formula according to the three selected benchmark individuals, the first updating formula is:

[0137] ,in, represents the characteristic value of the individual with mutation, , and They represent the eigenvalues ​​of the three reference individuals, including the position, roll angle, pitch angle and power of the first eigenvector. Indicates the scaling factor.

[0138] The above describes an embodiment of the present embodiment, but the present embodiment is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the guidance of the present embodiment, ordinary technicians in this field can also make many forms, all of which are within the protection of the present embodiment.

Claims

1. An unmanned ship control system integrating multiple communication links and interfaces, characterized in that: include: A data acquisition module, used for collecting environmental parameters, control parameters and state parameters of the unmanned ship during navigation within a first preset time period at a first preset time interval, and representing them through a first characteristic sequence; Environmental parameters include: water velocity, water direction, wave height, wave propagation direction, wind speed and wind direction; Control parameters include: speed, heading angle, power distribution; State parameters include: position, roll angle, pitch angle and power; A preprocessing and feature extraction module, used for processing the first feature sequence to obtain a second feature sequence, wherein the nth sequence unit of the second feature sequence represents the nth preprocessed environmental parameter, control parameter and state parameter; An unmanned ship state prediction module, used to predict the position, roll angle, pitch angle and power of the unmanned ship in a second preset time period according to a preset unmanned ship state prediction model, the input of the unmanned ship state prediction model is a second feature sequence, and the output is the position, roll angle, pitch angle and power of the unmanned ship in the second preset time period; Among them, the Pareto principle is used to improve the long short-term memory network, and an unmanned ship state prediction model is constructed based on the improved long short-term memory network. The unmanned ship state prediction model includes: input feature screening layer, grouping feature processing layer, feature fusion layer and output layer; The input feature screening layer is used to divide the features of the sequence units in the second feature sequence into key features and secondary features according to the Pareto principle, and obtain the key feature sequence and the secondary feature sequence respectively. The specific steps include: Step 301, calculating the Pearson correlation coefficient score between each feature of the sequence unit in the second feature sequence and the state parameter; Step 302, arranging the calculation results of step 301 in descending order; Step 303, selecting the first K features sorted in step 302 as key features, and the remaining features as secondary features, and finally obtaining a key feature sequence and a secondary feature sequence; The grouping feature processing layer is used to process the key feature sequence and the secondary feature sequence respectively; The feature fusion layer is used to splice the processing results of the key feature sequence and the secondary feature sequence to obtain the global feature; The output layer is used to input the global features into the first classifier, the second classifier, the third classifier and the fourth classifier respectively, and the classification spaces of the first classifier, the second classifier, the third classifier and the fourth classifier respectively represent the position, the roll angle, the pitch angle and the power; The control parameter optimization module is used to optimize the control parameters of the unmanned ship according to the improved genetic algorithm to obtain the optimal control parameters.

2. The unmanned ship control system integrating multiple communication links and interfaces according to claim 1, characterized in that: The water flow direction, wind direction and heading angle are doubly encoded using sine and cosine functions, the roll angle and pitch angle are represented by real number encoding, the power distribution is represented by the power ratio of the left and right thrusters, and the position is represented by latitude and longitude coordinates.

3. The unmanned ship control system integrating multiple communication links and interfaces according to claim 1, characterized in that: The first feature sequence is processed to obtain a second feature sequence, and the specific steps include: Step 201: extract wind flow disturbance synthesis features according to the first feature sequence, and concatenate them after the first feature sequence. The calculation formula of wind flow disturbance synthesis features is: ,in, represents the synthetic characteristics of wind flow disturbance, Indicates the water flow velocity, Indicates the direction of water flow. Indicates wind speed, Indicates wind direction. represents the angle between the water flow and the wind direction, and t represents the first preset time interval; Step 202: extract the wind-wave coupling power consumption feature according to the first feature sequence, and concatenate it after the first feature sequence. The calculation formula of the wind-wave coupling power consumption feature is: ,in, represents the wind-wave coupling power consumption characteristics, represents the density of water, Indicates wave height, It represents the angle between wind direction and wave height propagation direction. It represents the angle between the water flow direction and the wave height propagation direction. Indicates the difference between the wave height at the current time point and the wave height at the previous time point; Step 203, extract the comprehensive feature of speed and heading angle according to the first feature sequence, and splice it after the first feature sequence. The calculation formula of the comprehensive feature of speed and heading angle is: ,in, It represents the comprehensive characteristics of speed and heading angle. Indicates the speed. represents the heading angle, Indicates the difference between the speed at the current time point and the speed at the previous time point; Step 204: extract the dynamic stability feature of the hull according to the first feature sequence and concatenate it after the first feature sequence. The calculation formula of the dynamic stability feature of the hull is: ,in, represents the dynamic stability characteristics of the hull, N represents the number of time points, a, b, c and d represent the indexes of the time points, and denote the roll angles at the ath time point and the bth time point, respectively. and denote the pitch angles at the cth time point and the dth time point, respectively. Indicates the difference between the roll angle at the current time point and the roll angle at the previous time point. Indicates the difference between the pitch angle at the current time point and the pitch angle at the previous time point. represents the pitch angle at the ath time point; Step 205: Use the z-score normalization method to normalize the concatenated first feature sequence to obtain a second feature sequence.

4. The unmanned ship control system integrating multiple communication links and interfaces according to claim 1, characterized in that: The calculation formula of the Pearson correlation coefficient score in step 301 is: ； in, represents the Pearson correlation coefficient score of the e-th feature of the second feature sequence, The number of features representing the state parameters, represents the index of the feature of the state parameter, N represents the number of time points, f, g and h represent the time point index, and represent the feature values ​​of the e-th feature of the second feature sequence at the f-th time point and the g-th time point, respectively. and They represent the preset values ​​of the j-th feature of the state parameter at the f-th time point and the h-th time point respectively, where the preset values ​​represent predicted feature values ​​and are calculated based on a linear equation constructed using historical data.

5. The unmanned ship control system integrating multiple communication links and interfaces according to claim 1, characterized in that: The grouping feature processing layer is used to process the key feature sequence and the secondary feature sequence respectively. Specifically, the key feature sequence is processed by the first network layer to obtain a first hidden state, and the secondary feature sequence is processed by the second network layer to obtain a second hidden state; Among them, the calculation formula of the sth sequence unit of the key feature sequence of the first network layer includes: ； ； ； ； ； ； in, , , , , and They represent the input gate, forget gate, output gate, candidate memory unit, memory unit update and hidden state update of the sth sequence unit of the key feature sequence respectively. represents the sth sequence unit of the key feature sequence, and They represent the hidden states output by the s-th sequence unit and the s-1-th sequence unit, respectively. , and Respectively represent the first weight parameter, the first bias parameter and the second weight parameter of the input gate, represents the enhanced weight of the memory cell state of the previous sequence unit, Indicates that the sth sequence unit and the hidden state output by the s-1th sequence unit To splice, represents the memory cell update of the s-1th sequence unit, , and Respectively represent the first weight parameter, the first bias parameter and the second weight parameter of the forget gate, , and Respectively represent the first weight parameter, the first bias parameter and the second weight parameter of the output gate, and They represent the first weight parameter and the first bias parameter of the candidate memory unit respectively, sigmoid represents the sigmoid activation function, and tanh represents the hyperbolic tangent activation function; Extract the hidden state output by the Nth sequence unit as the first hidden state; The second network layer removes the input gate and output gate, candidate memory cells, and retains the forget gate, memory cell update, and hidden state update.

6. The unmanned ship control system integrating multiple communication links and interfaces according to claim 1, characterized in that: The control parameters of the unmanned ship are optimized according to the improved genetic algorithm. The specific steps include: Step 401, generating an initialization population including P individuals that meets the constraint conditions based on a spatial hierarchical method; Step 402, constructing an objective function according to the unmanned ship state prediction model; Step 403, calculating the fitness value of each individual in the population according to the objective function; Step 404, retaining V individuals with the highest fitness in the current population, and randomly selecting two individuals from them as parents; Step 405, generating a random number ranging from 0 to 1, and when the random number is less than a first preset threshold, performing a crossover operation that satisfies the constraint condition, wherein the crossover operation represents selecting an element value of one position of the first feature vector corresponding to two individuals in the parent generation to be exchanged, wherein the first threshold is a custom parameter; Step 406, updating the population individuals based on the differential mutation strategy; Step 407, determine whether the current number of iterations is greater than or equal to the maximum number of iterations, and use the individual with the largest fitness value in the initialized population as the optimal control parameter, otherwise return to step 403, where the maximum number of iterations is a custom parameter.

7. The unmanned ship control system integrating multiple communication links and interfaces according to claim 6, characterized in that: The objective function is calculated as follows: ； in, represents the fitness value of the objective function, , and represent the weight coefficients of the control distance target, control stability target and control energy consumption respectively, and , and The sum is 1, and They respectively represent the longitude of the current individual's unmanned ship position obtained by the unmanned ship state prediction model and the longitude of the individual's target position. and They respectively represent the latitude of the current individual's unmanned ship position obtained by the unmanned ship state prediction model and the latitude of the individual's target position. Represents a positive value between 0 and 1. represents the slope coefficient, and They represent the roll angle and pitch angle of the unmanned ship obtained by the current individual through the unmanned ship state prediction model, represents the weight coefficient of the pitch angle relative to the roll angle, It represents the power of the unmanned boat obtained by the current individual through the unmanned boat state prediction model.

8. The unmanned ship control system integrating multiple communication links and interfaces according to claim 6, characterized in that: The specific steps of step 406 include: Step 501, randomly select Y individuals from the population as individuals to be mutated; Step 502, randomly selecting three different benchmark individuals from the population; Step 503, update the mutated individuals using the first updating formula according to the three selected benchmark individuals. The first updating formula is: ,in, represents the characteristic value of the individual with mutation, , and They represent the eigenvalues ​​of the three reference individuals, including the position, roll angle, pitch angle and power of the first eigenvector. Indicates the scaling factor.

Images