Unmanned ship command and control system based on large language model

Abstract

The invention provides an unmanned ship command and control system based on a large language model, and the system comprises a situation information collection module which employs a camera, a radar, and a sonar sensor to form a multi-sensor array, carries out the collection of situation information, carries out the preprocessing of the collected situation information, and carries out the processing of the collected situation information. The strategy generation module is used for extracting characteristic indexes through a large language model and generating an auxiliary strategy set; according to the decision and execution module, an operation end receives the situation information, and an operator selects a confrontation strategy in combination with experience and converts the confrontation strategy into an instruction; the unmanned ship end analyzes the instruction, uses path planning and PID control to adjust navigation parameters, controls equipment, and feeds back an execution condition; and the communication module selects a UHF frequency band and a time division multiple access protocol, adopts an encryption standard algorithm to encrypt transmission data, and guarantees communication stability by monitoring an error rate. The system has the advantages of comprehensive and accurate information acquisition, deep information processing and analysis, reliable real-time communication, intelligent aid decision making and accurate task execution and feedback.

Description

Technical Field

[0001] The present invention relates to the technical field of unmanned boat control, and in particular to an unmanned boat command and control system based on a large language model. Background Art

[0002] In recent years, with the rapid advancement of science and technology, the use of unmanned aerial vehicles (UAVs) in the marine sector has become increasingly widespread. With advantages such as small size, low cost, and high maneuverability, UAVs play an important role in marine monitoring, resource exploration, and military reconnaissance. Currently, UAVs at sea primarily utilize an autonomous mission execution model, where operators pre-set mission instructions and the UAVs autonomously complete the mission according to those instructions. This model demonstrates high efficiency and reliability in some simple mission scenarios. For example, in marine environmental monitoring missions, UAVs can conduct water quality sampling and meteorological observations along pre-set routes, accurately acquiring relevant data. In military reconnaissance missions, UAVs can quietly approach target areas and gather intelligence. However, when faced with complex situations such as anti-smuggling, anti-stowaway operations, and other maritime confrontations, this autonomous mission execution model reveals significant limitations.

[0003] In maritime confrontation scenarios, the situation is even more complex. The enemy may resort to interference, attacks, and other means to counter the UAV. The UAV needs to understand the battlefield situation in real time and respond according to the operator's instructions. If the UAV cannot promptly transmit situational information to the operator, the operator will not be able to make accurate decisions, and the UAV will not be able to effectively perform the confrontation mission. In addition, in complex marine environments, UAVs may encounter various emergencies such as equipment failures and marine obstacles. In such cases, the UAV needs to communicate with the operator to determine the best solution. Therefore, in these complex mission scenarios, there is an urgent need for an UAV command and control system that can achieve human-machine integration. Summary of the Invention

[0004] In order to solve the technical problems mentioned in the current background technology, the purpose of the present invention is to provide an unmanned boat command and control system based on a large language model.

[0005] In order to achieve the above object, the present invention adopts the following technical solutions:

[0006] An unmanned boat command and control system based on a large language model, comprising:

[0007] The information collection module and the system include an operation terminal and an unmanned boat terminal; the first situation data of the target at sea is collected in real time, and the first situation data is pre-processed to obtain the second situation data, and the second situation data is transmitted to the operation terminal and the strategy generation module;

[0008] A strategy generation module inputs the second situation data into a preset large language model, extracts characteristic indicators, outputs an auxiliary strategy set based on the characteristic indicators, and transmits the auxiliary strategy set to the operation end;

[0009] The decision-making and execution module and the operator formulate a countermeasure strategy based on the second situation data and the auxiliary strategy set, and send the countermeasure strategy to the unmanned boat end; adjust the navigation parameters and equipment control parameters of the unmanned boat based on the countermeasure strategy, and the unmanned boat executes the countermeasure task based on the navigation parameters and equipment control parameters, and feeds back the execution result to the strategy generation module;

[0010] The communication module ensures the security and stability of data communication between the operating end and the unmanned boat end through UHF frequency band, time division multiple access transmission and AES algorithm, and compresses the transmitted data.

[0011] Furthermore, the first situation data includes image data, radar data and sonar data;

[0012] Based on physical characteristics, a multi-sensor array including wide-angle cameras, multi-band radars, and sonars is deployed to collect first-hand situational data in real time;

[0013] The image data is collected and acquired by a wide-angle camera device based on the optical imaging principle;

[0014] The radar data is acquired by a multi-band radar through electromagnetic wave reflection characteristics and Doppler effect detection;

[0015] The sonar data is acquired by a sonar device using a detection mode combining active and passive detection.

[0016] Furthermore, the pretreatment includes a first-stage treatment and a second-stage treatment;

[0017] The first level of processing is to remove noise and interference in the image data by using adaptive median filtering technology, and the specific steps are as follows:

[0018] 1) Set the filter window and calculate the pixel values ​​within the filter window to obtain the maximum pixel value Z max , minimum value Z min and median Z med ;

[0019] 2) Before making noise judgment, Z must be satisfied. min <Z med <Z max :

[0020] When Z min <I(x,y)<Z max When , output I(x,y), otherwise output Z med, I(x,y) is the original pixel value with coordinates (x,y) in the image data;

[0021] When Z med ≤Z min or Z med ≥Z max When Z is satisfied, the filter window is expanded until Z min <Z med <Z max Or the maximum value preset in the filtering window is reached.

[0022] Furthermore, the second level of processing refers to removing the radar data and sonar data through Wiener filtering to perform noise reduction processing, and the specific steps are as follows:

[0023] 1) Calculating the observation signals of the multi-band radar and sonar, and inputting the observation signals into a preset Wiener filter:

[0024] y(n)=x(n)+v(n)

[0025] Where x(n) is the true signal at the current time n, and v(n) is the additive noise at the current time n;

[0026] 2) The Wiener filter outputs the optimal estimate x(n) of the real signal x(n) of the multi-band radar and sonar:

[0027]

[0028] Wherein, h(i) is the i-th order coefficient of the Wiener filter, y(nt) is the observation signal with a delay time of t, and M is the total order of the filter;

[0029] By minimizing the mean square error E[(x(n)-x(n)) 2 ] to calculate h(i);

[0030] 3) Solve the Wiener-Hoff equation:

[0031]

[0032] Among them, R yy (l) is the autocorrelation function of the observed signal y(n), l is the time delay;

[0033] Automatically suppress R by the filter coefficient h(i) yy The noise-dominated component in (l).

[0034] Furthermore, the large language model includes an input layer, a hidden layer, a feature extraction layer, a processing layer and an output layer;

[0035] The input layer inputs the second situation data;

[0036] The hidden layer output is:

[0037] H=σ(W·X+b)

[0038] Where W is the weight matrix of the preset large language model, b is the preset bias vector, and σ is the activation function;

[0039] The hidden layer output H refers to the high-dimensional nonlinear transformation result of the data, which is used to enhance feature extraction.

[0040] Furthermore, the feature extraction layer extracts feature indicators including image feature indicators, radar feature indicators and sonar feature indicators based on the second situation data;

[0041] Encode the feature indices into a unified feature index vector X:

[0042] X=[x1,x2,…,x n ]

[0043] Among them, x n is the n-th dimension eigenvalue;

[0044] The processing layer performs target threat quantitative assessment and unmanned boat capability assessment based on a preset expert experience database and characteristic indicators;

[0045] The calculation formula of the target threat quantitative assessment T is as follows:

[0046]

[0047] Among them, D α The threat level of the characteristic indicator described in item α, Q α is the value of the characteristic index described in item α;

[0048] The calculation formula of the unmanned boat capability assessment C is as follows:

[0049]

[0050] Among them, R k is the preset k-th capability index weight of the unmanned boat, c k is the measured value of the kth capability indicator of the unmanned boat.

[0051] Further, generating a situation score E based on the target threat quantitative assessment and the unmanned boat capability assessment;

[0052] Generate auxiliary strategy set P={p1,p2,…,p m}, each policy p m There is a corresponding evaluation score A(p m );

[0053] The output layer outputs the auxiliary policy set.

[0054] Furthermore, the operator at the operation end selects a countermeasure strategy from the auxiliary strategy set P based on his / her own experience, professional knowledge and the second situation data, and adjusts the evaluation score of the strategy in the auxiliary strategy set based on different mission scenarios and goals;

[0055] The countermeasure strategy is converted into instruction I and transmitted to the unmanned boat.

[0056] Furthermore, the unmanned boat encodes the instruction in a standardized encoding format and sets the instruction set to I = {i1, i2, ..., i n};

[0057] Adjusting navigation parameters and equipment control parameters of the unmanned boat based on the instruction set;

[0058] Monitor task execution in real time and feed back execution results to the strategy generation module.

[0059] Furthermore, the UHF band implements electromagnetic spectrum planning by defining the operating frequency range;

[0060] Dividing the communication time into a number of time slots by the time division multiple access (TDMA) protocol;

[0061] The unmanned boat end and the operating end are respectively allocated specific time slots for information transmission;

[0062] Encrypting the transmitted data using the AES algorithm;

[0063] Evaluate communication quality by monitoring communication status in real time and calculating bit error rate;

[0064] When the bit error rate reaches a preset first threshold, it is determined to be a communication failure;

[0065] When the communication failure occurs, a retransmission mechanism and communication frequency band switching measures are adopted;

[0066] The system first starts the retransmission mechanism to retransmit the data. When the number of retransmissions reaches a preset second threshold, the communication frequency band switching measure is initiated.

[0067] In order to ensure real-time data transmission, data compression technology is used to compress the transmitted data.

[0068] Compared with the prior art, the advantages of the present invention are:

[0069] 1. This invention uses a multi-sensor array to collaboratively collect visible light, radar, and sonar data, and employs a combination of adaptive filtering and model-driven analysis. This overcomes the limitations of traditional systems that independently process multimodal data. The large language model uses nonlinear transformations between feature extraction and hidden layers to uniformly encode heterogeneous features such as image contours, target motion parameters, and underwater distance into high-dimensional vectors. This significantly improves the depth of situational information fusion and the accuracy of analysis, providing a comprehensive and consistent foundation for environmental cognition in complex maritime confrontation scenarios.

[0070] 2. This invention innovatively introduces a large language model-driven strategy generation module, which implements quantitative target threat assessment and dynamic analysis of UAV capabilities based on an expert experience database, generating an interpretable auxiliary strategy set. Operators can optimize strategies based on real-time situation data and model recommendations. This approach retains the flexibility and empirical value of human decision-making while also compensating for the response lag of purely manual decision-making in complex scenarios through the model's situation scoring and strategy evaluation functions, thereby improving the adaptability and real-time performance of the adversarial strategy.

[0071] 3. Through the collaborative design of UHF band communication, time division multiple access protocol and dynamic encryption technology, the present invention establishes a highly robust communication link in the complex electromagnetic environment at sea. The communication module realizes fault self-diagnosis based on bit error rate monitoring, and combines the retransmission mechanism with the frequency band switching strategy to effectively resist multipath fading and malicious interference. At the same time, the decision-making execution module ensures the accurate execution and real-time feedback of the countermeasure strategy through standardized instruction analysis, dynamic path planning and equipment closed-loop control, forming a complete closed loop, thereby improving the system's task completion rate and anti-damage capability. BRIEF DESCRIPTION OF THE DRAWINGS

[0072] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0073] Figure 1 Schematic diagram of the system workflow of the present invention;

[0074] Figure 2 Schematic diagram of the flow of the large language model of the present invention;

[0075] Figure 3 This is a schematic diagram of situation data collection of the present invention. DETAILED DESCRIPTION

[0076] To achieve the above objectives, the present invention is implemented through the following technical solutions. The present invention provides an unmanned boat command and control system based on a large language model. Figures 1 to 3 , the system comprises:

[0077] The information collection module and the system include an operation terminal and an unmanned boat terminal; the first situation data of the target at sea is collected in real time, and the first situation data is pre-processed to obtain the second situation data, and the second situation data is transmitted to the operation terminal and the strategy generation module;

[0078] The first situational data includes image data, radar data, and sonar data.

[0079] Based on physical characteristics, a multi-sensor array including wide-angle cameras, multi-band radars, and sonars is deployed to collect first-hand situational data in real time;

[0080] The wide-angle camera obtains image data based on the principle of optical imaging;

[0081] Multi-band radar uses electromagnetic wave reflection and the Doppler effect to detect targets and obtain radar data;

[0082] Sonar uses a combination of active and passive methods to detect underwater targets and obtain sonar data;

[0083] To enhance the accuracy and comprehensiveness of situational data collection, sensors are strategically positioned throughout the unmanned vessel. The camera is mounted at least three meters above the deck to provide a wider field of view; the radar is installed on the top of the vessel to minimize obstruction; and the sonar is mounted on the bottom to facilitate underwater target detection. The sensors work collaboratively. For example, when the radar detects a target, it can direct the camera to focus on it, improving target recognition accuracy.

[0084] Preprocessing the first situation data to obtain second situation data includes first-level processing and second-level processing;

[0085] The first level of processing is to pre-process the image data collected in real time by the wide-angle camera. Adaptive median filtering technology is used to effectively eliminate image noise caused by wave reflections and rain and fog interference, and improve the clarity of the target outline. The specific steps are as follows:

[0086] 1) Set the filter window and calculate the pixels within the filter window to get the maximum value Z max , minimum value Z min and median Z med ;

[0087] 2) Noise judgment rules: Z must be satisfied before making noise judgment min <Z med <Z max :

[0088] When Z min <I(x,y)<Z maxWhen , output I(x,y), otherwise output Z med , I(x,y) is the original pixel value of the coordinate (x,y) in the visible light image;

[0089] When Z med ≤Z min or Z med ≥Z max When Z min <Z med <Z max Or k = k max (k max =3);

[0090] The final output is the denoised image O(x,y).

[0091] The second stage pre-processes the radar and sonar data, suppresses sea clutter, equipment electromagnetic interference and other noise based on Wiener filtering, and extracts accurate target distance and speed parameters. The specific steps are as follows:

[0092] 1) Calculate multi-band radar and sonar observation signals and input the observation signals into the preset Wiener filter:

[0093] y(n)=x(n)+v(n)

[0094] Where x(n) is the true signal at the current time n, and v(n) is the additive noise at the current time n;

[0095] 2) The Wiener filter outputs the optimal estimate of the true signal x(n)

[0096]

[0097] Where h(i) is the i-th order coefficient of the Wiener filter, y(nt) is the observed signal with a delay of time t, and M is the total order of the filter;

[0098] Calculate h(i) by minimizing the mean square error;

[0099] 3) Solve the Wiener-Hoff equation:

[0100]

[0101] Among them, R yy (l) is the autocorrelation function of the observed signal y(n), l is the time delay;

[0102] Automatically suppress R by the filter coefficient h(i) yy (l) The noise-dominated component;

[0103] A strategy generation module inputs the situation data into a preset large language model, extracts characteristic indicators, outputs an auxiliary strategy set based on the characteristic indicators, and transmits the auxiliary strategy set to the operation end;

[0104] The large language model is trained using historical situation data, professional knowledge base, historical adversarial strategies, and the execution results of historical adversarial strategies.

[0105] The large language model includes input layer, hidden layer, feature extraction layer, processing layer and output layer;

[0106] The input layer inputs the second situation data;

[0107] The output of the hidden layer is:

[0108] H=σ(W·X+b)

[0109] Where W is the model weight matrix, obtained through massive historical adversarial scene training, b is the bias vector, which optimizes the model fitting ability, and σ is the activation function, such as the ReLU activation function σ(z) = max(0,z), which enhances the ability of nonlinear feature extraction;

[0110] The feature extraction layer extracts feature indicators including image feature indicators, radar feature indicators, and sonar feature indicators based on the second situation data. The specific steps are as follows:

[0111] 1) Image feature index extraction,

[0112] For the image data after noise reduction, calculate the contour feature vector F of the target in the image data;

[0113] Similarity matching S is performed based on the target contour feature vector F and the preset ship feature library i , the technical formula is as follows:

[0114]

[0115] Among them, F d is the contour feature vector in the ship feature library;

[0116] Select similarity matching S i The highest one is used as the image feature index;

[0117] 2) Radar feature index extraction based on radar data and Doppler effect formula, the formula is as follows:

[0118]

[0119] Among them, f d is the measured Doppler shift, λ is the wavelength of the transmitted electromagnetic wave, and θ is the angle between the target motion direction and the radar beam;

[0120] Extract radar characteristic indicators including the target radial velocity v and motion direction θ;

[0121] 3) Extract sonar feature indicators by using the acoustic wave delay formula:

[0122]

[0123] Among them, t s is the active sonar echo delay, v s is the speed of sound wave propagation in water, d s is the underwater target distance;

[0124] Extract sonar feature indicators based on underwater target distance;

[0125] Encode the feature indices into a unified feature index vector X:

[0126] X=[x1,x2,…,x n ]

[0127] Among them, x n is the n-th dimension eigenvalue;

[0128] The processing layer conducts quantitative assessment of target threats and UAV capability based on the preset expert experience database and characteristic indicators;

[0129] The target threat quantitative assessment formula is as follows:

[0130]

[0131] Among them, D α The threat level of the αth characteristic indicator, Q α is the value of the αth characteristic index, and N is the total number of characteristic indicators.

[0132] Our unmanned boat capability assessment C is calculated as follows:

[0133]

[0134] Among them, R k is the preset k-th capability index weight of the unmanned boat, c k is the measured value of the kth capability indicator of the unmanned boat;

[0135] In this embodiment, the capabilities of the unmanned boat include maximum speed, weapon range and accuracy, stealth performance, sensor detection range, and anti-interference capability;

[0136] The situation score E is generated based on the quantitative assessment of target threat and the assessment of the capability of the unmanned boat. The formula is as follows:

[0137] E=f(T,C)

[0138] Among them, f is a nonlinear evaluation function implemented by the large language model, which comprehensively weighs the advantages and disadvantages of the enemy and our own situation;

[0139] Generate auxiliary strategy set P={p1,p2,…,p m}, each policy p m There is a corresponding evaluation score A(p m ), the formula is as follows:

[0140] A(p m )=g(E,p m )

[0141] Among them, g is a function based on the large language model, based on factors such as strategy feasibility and effectiveness;

[0142] The output layer outputs the auxiliary strategy set;

[0143] The decision-making and execution module and the operator formulate a countermeasure strategy based on the second situation data and the auxiliary strategy set, and send the countermeasure strategy to the unmanned boat end; adjust the navigation parameters of the unmanned boat based on the countermeasure strategy, and the unmanned boat executes the countermeasure task based on the navigation parameters and feeds back the execution result to the strategy generation module;

[0144] The operator selects a countermeasure strategy from the auxiliary strategy set P based on his or her own experience, professional knowledge, and the second situation data. The operator can also adjust the evaluation score of the strategy in the auxiliary strategy set according to different mission scenarios and goals.

[0145] The confrontation strategy is converted into instruction I, which contains the specific operations performed by the unmanned boat, such as navigation route, speed, equipment use, etc., and the instruction I is transmitted to the unmanned boat end.

[0146] After receiving the command, the unmanned boat first parses the command;

[0147] In this embodiment, the instructions are encoded in a standardized format, such as JSON or XML, and the instruction set is set to I = {i1, i2, ..., i n}, through regular expression matching or parser, the command is converted into a format that can be processed by the unmanned boat. For example, the command to control the navigation route is parsed to obtain the target coordinates (x t ,y t ) and the desired velocity v d .

[0148] Based on the instruction set, adjust the navigation parameters and equipment control parameters of the unmanned boat as follows:

[0149] Navigation parameters include navigation route adjustment and speed adjustment;

[0150] The navigation route adjustment uses a path planning algorithm to calculate the optimal obstacle avoidance path;

[0151] Speed ​​adjustment according to the desired speed v d and current speed v c Speed ​​adjustment is performed using PID control algorithm.

[0152] Equipment control includes weapon system control and jamming equipment control;

[0153] In this embodiment, if the weapon system controls the command to launch a missile, the launch angle θ needs to be calculated using the following formula:

[0154]

[0155] Among them, x t with y t is the target coordinate, x c with y c is the coordinate of the unmanned boat itself, v t is the target movement speed (unit: m / s), v m The maximum speed supported by the unmanned boat power system (unit: m / s), It is the angle between the target's moving direction and the line connecting the unmanned boat and the target.

[0156] In this embodiment, the jammer device controls, if the instruction requires electronic jammer to be used on the target,

[0157] According to the electromagnetic characteristics and position of the target, the interference power P is adjusted. The interference power is adjusted according to the target distance d to meet the requirements.

[0158]

[0159] Where P0 is the initial power and d0 is the initial distance.

[0160] The execution module monitors mission execution in real time and provides feedback to the strategy generation module. This feedback includes the UAV's current position, speed, and device status. Sensors collect this data, which is then processed and sent to the information processing module in the form of data packets for subsequent adjustments and optimization.

[0161] The communication module uses UHF frequency band, time division multiple access transmission and AES algorithm to ensure the security and stability of data communication between the operator and the unmanned boat.

[0162] Based on the complex electromagnetic environment and multipath fading at sea, a frequency band within the 300MHz-3GHz range is selected. This frequency band has good propagation characteristics at sea and can reduce the impact of obstacles.

[0163] The time is divided into several time slots through the time division multiple access (TDMA) protocol;

[0164] The unmanned boat end and the operating end are respectively allocated specific time slots for information transmission to avoid communication conflicts and improve communication efficiency.

[0165] To prevent information transmission from being eavesdropped and interfered with, the encryption standard (AES) algorithm is used for encryption.

[0166] The encryption formula is:

[0167] G=AES(h,J)

[0168] Where G is the encrypted ciphertext, h is the plaintext information to be encrypted, and J is the encryption key;

[0169] The AES algorithm is highly secure and efficient and can effectively protect the confidentiality of communication information.

[0170] Monitor the communication status in real time and evaluate the communication quality by calculating the bit error rate (BER).

[0171] The bit error rate BER is the number of received error bits L e The total number of bits transmitted L t The ratio is as follows:

[0172]

[0173] When the bit error rate exceeds a preset first threshold, it is determined that a communication failure occurs.

[0174] When a failure occurs, a retransmission mechanism and switching of the communication frequency band are adopted. The number of retransmissions is set to o and the time interval between each retransmission is t. r .

[0175] If retransmission fails after o times, it will automatically switch to the backup communication band. The backup band is selected based on the band availability and interference situation. By monitoring the signal strength and interference level of each band in real time, the band with the strongest signal and the least interference is selected for communication.

[0176] To ensure real-time transmission of information, data compression technology is used to compress situational information; by improving the compression ratio, the data transmission volume is reduced, the transmission time is shortened, and real-time performance is improved. The compression ratio is the ratio of the original data size to the compressed data size.

[0177] In summary, the present invention solves the noise interference problem in complex marine environments through the heterogeneous sensor collaboration mechanism and two-stage adaptive filtering technology; combined with the feature extraction and situation assessment capabilities of the large language model, it realizes cross-domain fusion cognition of sea, air and underwater targets, significantly improving the comprehensiveness of environmental perception and the depth of analysis.

[0178] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. An unmanned boat command and control system based on a large language model, characterized in that: include: The information collection module and system include the operation terminal and the unmanned boat terminal; Collect first situation data of the target at sea in real time, pre-process the first situation data to obtain second situation data, and transmit the second situation data to the operation terminal and the strategy generation module; A strategy generation module inputs the second situation data into a preset large language model, extracts characteristic indicators, outputs an auxiliary strategy set based on the characteristic indicators, and transmits the auxiliary strategy set to the operation end; The decision-making and execution module and the operator formulate a countermeasure strategy based on the second situation data and the auxiliary strategy set, and send the countermeasure strategy to the unmanned boat end; adjust the navigation parameters and equipment control parameters of the unmanned boat based on the countermeasure strategy, and the unmanned boat executes the countermeasure task based on the navigation parameters and equipment control parameters, and feeds back the execution result to the strategy generation module; The communication module ensures the security and stability of data communication between the operating end and the unmanned boat end through UHF frequency band, time division multiple access transmission and AES algorithm, and compresses the transmitted data.

2. The system according to claim 1, wherein: The first situation data includes image data, radar data and sonar data; Based on physical characteristics, a multi-sensor array including wide-angle cameras, multi-band radars, and sonars is deployed to collect first-hand situational data in real time; The image data is collected and acquired by a wide-angle camera device based on the optical imaging principle; The radar data is acquired by a multi-band radar through electromagnetic wave reflection characteristics and Doppler effect detection; The sonar data is acquired by a sonar device using a detection mode combining active and passive detection.

3. The system according to claim 1, wherein: The pretreatment includes a first-stage treatment and a second-stage treatment; The first level of processing is to remove noise and interference in the image data by using adaptive median filtering technology, and the specific steps are as follows: 1) Set the filter window and calculate the pixel values ​​within the filter window to obtain the maximum pixel value Z max , minimum value Z min and median Z med ; 2) Before making noise judgment, Z must be satisfied. min <Z med <Z max : When Z min <I(x,y)<Z max When , output I(x,y), otherwise output Z med , I(x,y) is the original pixel value with coordinates (x,y) in the image data; When Z med ≤Z min or Z med ≥Z max When Z is satisfied, the filter window is expanded until Z min <Z med <Z max Or the maximum value preset in the filtering window is reached.

4. The system according to claim 3, characterized in that The second level of processing involves removing the noise from the radar and sonar data through Wiener filtering. The specific steps are as follows: 1) Calculating the observation signals of the multi-band radar and sonar, and inputting the observation signals into a preset Wiener filter: y(n)=x(n)+v(n) Where x(n) is the true signal at the current time n, and v(n) is the additive noise at the current time n; 2) The Wiener filter outputs the optimal estimate of the true signal x(n) of the multi-band radar and sonar Wherein, h(i) is the i-th order coefficient of the Wiener filter, y(nt) is the observation signal with a delay time of t, and M is the total order of the filter; By minimizing the mean square error To calculate h(i); 3) Solve the Wiener-Hoff equation: Among them, R yy (l) is the autocorrelation function of the observed signal y(n), l is the time delay; Automatically suppress R by the filter coefficient h(i) yy The noise-dominated component in (l).

5. The system according to claim 1, wherein: The large language model includes an input layer, a hidden layer, a feature extraction layer, a processing layer and an output layer; The input layer inputs the second situation data; The hidden layer output is: H=σ(W·X+b) Where W is the weight matrix of the preset large language model, b is the preset bias vector, and σ is the activation function; The hidden layer output H refers to the high-dimensional nonlinear transformation result of the data, which is used to enhance feature extraction.

6. The system according to claim 5, characterized in that The feature extraction layer extracts feature indicators including image feature indicators, radar feature indicators and sonar feature indicators based on the second situation data; Encode the feature indices into a unified feature index vector X: X=[x1,x2,…,x n ] Among them, x n is the n-th dimension eigenvalue; The processing layer performs target threat quantitative assessment and unmanned boat capability assessment based on a preset expert experience database and characteristic indicators; The calculation formula of the target threat quantitative assessment T is as follows: Among them, D α The threat level of the characteristic indicator described in item α, Q α is the value of the characteristic index described in item α; The calculation formula of the unmanned boat capability assessment C is as follows: Among them, R k is the preset k-th capability index weight of the unmanned boat, c k is the measured value of the kth capability indicator of the unmanned boat.

7. The system according to claim 6, characterized in that Generating a situation score E based on the target threat quantitative assessment and the unmanned boat capability assessment; Generate auxiliary strategy set P={p1,p2,…,p m }, each policy p m There is a corresponding evaluation score A(p m ); The output layer outputs the auxiliary policy set.

8. The system according to claim 1, wherein: The operator at the operation end selects a countermeasure strategy from the auxiliary strategy set P based on his / her own experience, professional knowledge and the second situation data, and adjusts the evaluation score of the strategy in the auxiliary strategy set based on different mission scenarios and goals; The countermeasure strategy is converted into instruction I and transmitted to the unmanned boat.

9. The system according to claim 8, characterized in that The unmanned boat end encodes the instruction using a standardized encoding format and sets the instruction set to I = {i1, i2, ..., i n }; Adjusting navigation parameters and equipment control parameters of the unmanned boat based on the instruction set; Monitor task execution in real time and feed back execution results to the strategy generation module.

10. The system according to claim 1, wherein: The UHF frequency band realizes electromagnetic spectrum planning by defining the working frequency domain range; Dividing the communication time into a number of time slots by the time division multiple access (TDMA) protocol; The unmanned boat end and the operating end are respectively allocated specific time slots for information transmission; Encrypting the transmitted data using the AES algorithm; Evaluate communication quality by monitoring communication status in real time and calculating bit error rate; When the bit error rate reaches a preset first threshold, it is determined to be a communication failure; When the communication failure occurs, a retransmission mechanism and communication frequency band switching measures are adopted; The system first starts the retransmission mechanism to retransmit the data. When the number of retransmissions reaches a preset second threshold, the communication frequency band switching measure is initiated. In order to ensure real-time data transmission, data compression technology is used to compress the transmitted data.

Images