Large-scale unmanned ship scheduling method, system, equipment and medium

Abstract

The invention discloses a large-scale unmanned ship scheduling method, system, equipment and medium, and relates to the technical field of unmanned ship scheduling, and the method comprises the steps: building a disturbance prediction model based on a propulsion angle and a power parameter, and mapping propulsion control data through a nonlinear characteristic function; analyzing a stripping residual error according to sensor data to obtain a stripped attitude signal, and compressing abnormal disturbance; and constructing a model drift index according to the error window fluctuation, performing drift judgment based on the model drift index, and triggering model self-correction. According to the method, high-precision disturbance estimation is provided, the adaptability of the model under the dynamic water area condition is improved, the response capability of the system to the environment change is improved, the more stable and reliable unmanned ship scheduling capability is provided in the dynamically changing water area environment, scheduling errors caused by model drifting are avoided, and the scheduling efficiency is improved. And long-term stable operation of the unmanned ship dispatching system in a complex and changing environment is ensured.

Description

Technical Field

[0001] The present invention relates to the technical field of unmanned boat scheduling, and specifically to a large-scale unmanned boat scheduling method, system, device and medium. Background Art

[0002] In recent years, the technology of unmanned boats has developed rapidly and has been widely applied in fields such as surface cruising, environmental monitoring, and data collection. With the progress of intelligent control technology, the scheduling and control system based on sensor data and real-time feedback has become the core of the unmanned boat control system. The unmanned boat scheduling system obtains control parameters such as propulsion angle, power, and acceleration, and combines complex environmental conditions to perform real-time status monitoring and dynamic scheduling. In this process, the disturbance prediction model has become one of the key technologies. By accurately predicting the propulsion control data, it can effectively respond to the dynamic changes in the water surface environment. Existing unmanned boat scheduling methods have made certain progress in dealing with disturbances and dynamic errors in complex water environments, but there are still certain technical bottlenecks in the face of the demand for large-scale unmanned boat cluster scheduling.

[0003] Although existing unmanned boat scheduling technologies have achieved certain results in some specific applications, there are still deficiencies, especially in large-scale unmanned boat scheduling, accurate disturbance prediction, and error self-correction. Most existing disturbance prediction models are based on linear relationships for modeling, ignoring the non-linear characteristics between parameters such as propulsion angle and power in actual operations, resulting in low prediction accuracy. For complex disturbances in the water surface environment, traditional models are difficult to fully consider non-linear factors such as water flow changes and wind speed changes. Therefore, the prediction ability in a highly dynamic environment is particularly insufficient. Most existing technologies rely solely on a single control feedback mechanism in the processing of attitude signals, making it difficult to accurately separate interference signals caused by equipment operation and environmental factors, resulting in abnormal disturbances during attitude adjustment being difficult to compress and optimize, thus affecting the overall scheduling accuracy. Traditional model drift monitoring methods often only judge model drift based on the simple fluctuation of error values, and fail to adopt a more accurate dynamic feedback mechanism. Due to the lag of drift determination, the model fails to be corrected in time when drift occurs, affecting the stability and efficient scheduling ability of the unmanned boat. In view of the above deficiencies, the present invention proposes a non-linear disturbance prediction model based on propulsion angle and power parameters, which calculates and separates residuals through sensor data and compresses abnormal disturbances, thereby improving prediction accuracy and scheduling efficiency. By constructing a model drift index and performing drift judgment, the present invention can trigger a self-correction mechanism in real time when the model drifts, greatly improving the stability and adaptability of the system, and overcoming the problems that existing technologies cannot effectively handle complex dynamic disturbances and model drift during large-scale scheduling. Summary of the Invention

[0004] In view of the above existing problems, the present invention is proposed.

[0005] Therefore, the technical problem solved by the present invention is that the existing unmanned boat scheduling technology has low disturbance prediction accuracy, inaccurate attitude signal processing, lag in model drift judgment, and the problem of how to improve the accuracy and stability of large-scale unmanned boat scheduling.

[0006] To solve the above technical problems, the present invention provides the following technical solutions: A large-scale unmanned boat scheduling method, including constructing a disturbance prediction model based on propulsion angle and power parameters, and using a non-linear characteristic function to map propulsion control data; analyzing and peeling residuals according to sensor data to obtain the peeled attitude signal, and compressing abnormal disturbances; constructing a model drift index according to the error window fluctuation, and performing drift judgment and triggering model self-calibration based on the model drift index; the drift judgment includes continuously monitoring the change of the model error, dynamically evaluating the fluctuation of the error, analyzing the error in the current time period, accumulating the deviation and calculating the fluctuation amplitude, and using the average value and the front and back change difference of the error to construct a disturbance prediction model drift index to help judge whether there is a systematic error in the disturbance prediction model, and dynamically adjusting the time span and update frequency of the sliding window according to the task cycle; triggering the model self-calibration includes performing drift judgment based on the model drift index, taking different processing methods according to different drift thresholds, performing error verification and response testing after the model is updated to determine that the updated model runs stably, and if not up to standard, rolling back to the previous version.

[0007] As a preferred scheme of the large-scale unmanned boat scheduling method described in the present invention, wherein: the construction of the disturbance prediction model includes collecting input control parameters such as the propulsion direction angle, propeller power, current water depth, and platform acceleration of the unmanned boat as input factors for disturbance prediction, and establishing a disturbance response model to predict the attitude change of the platform.

[0008] Adopt a response construction method with non-linear characteristics, assign different weights to various input factors to measure the influence degree on the platform attitude change, obtain the adjustment coefficient through historical data learning, and perform normalization processing on the input, and present the disturbance prediction result through the attitude angle change.

[0009] As a preferred scheme of the large-scale unmanned boat scheduling method described in the present invention, wherein: the use of the non-linear characteristic function to map the propulsion control data includes fusing and mapping the control parameters including propulsion angle, propulsion power, acceleration, and depth in a non-linear relationship to form a function structure for predicting disturbances, and estimating the possible attitude response of the platform through the dynamic weights of different input parameters.

[0010] The fusion mapping includes judging the possibility of platform deflection by evaluating the angle between the propulsion direction and the platform heading through the system, determining the destructive impact of the propulsion power on the platform stability within a certain range, and judging whether there are sharp turn and sudden acceleration dynamic instability events in combination with the acceleration change trend.

[0011] Construct the amplification effect of depth on perturbation as a non-linear enhancement structure to simulate the amplification effect of water depth environment on platform control response.

[0012] Reflect the randomness and uncertainty of the perturbation through the fusion perturbation response deviation control mechanism. When the perturbation prediction model runs, obtain the real-time perturbation output according to the specific parameter combination, and perform classification processing by setting the response interval.

[0013] As a preferred solution of the large-scale unmanned boat scheduling method described in the present invention, wherein: the peeling off of residuals according to the sensor data analysis includes that the platform collects the actual attitude change angle in real time through the sensor, compares it with the predicted perturbation result provided by the perturbation prediction model, and analyzes and peels off the residuals.

[0014] As a preferred solution of the large-scale unmanned boat scheduling method described in the present invention, wherein: the compression of abnormal perturbations includes performing segmented compression calculation according to the preset error level standard.

[0015] Combining the compression mechanism and the error, use an increasing function to execute the decreasing compression rate. In the final output signal, retain the real water body perturbation part and actively peel off the high-amplitude error generated by abnormal device operation.

[0016] As a preferred solution of the large-scale unmanned boat scheduling method described in the present invention, wherein: the construction of the model drift index includes continuously monitoring the change of the model error by adopting a sliding time window mechanism, and constructing the model drift index by evaluating the cumulative deviation and fluctuation amplitude of the error within the current time period.

[0017] According to the error average value and the difference between the current error and the previous error, and perform integral processing based on a certain time interval to capture whether there is a systematic error accumulation phenomenon in the perturbation prediction model. When a new error data point is detected during operation, incorporate the new error data point into the current sliding window, compare the error with the previous time period, and update the drift index.

[0018] Dynamically adjust the time span and update frequency of the sliding window according to the task cycle.

[0019] As a preferred solution of the large-scale unmanned boat scheduling method described in the present invention, wherein: the triggering of the model self-calibration includes performing drift judgment based on the model drift index and adopting different processing methods according to different drift thresholds.

[0020] If the existing disturbance prediction model structure does not adapt to the current changes in disturbance characteristics, the training module is called to reload the recently collected control input and platform response data, update the fitting parameters and adjust the model structure, and construct a new round of disturbance prediction models.

[0021] After the disturbance prediction model is updated, error verification and response testing are carried out to confirm whether the prediction error returns to stability. If the standard is not met, roll back to the previous version and continue to run.

[0022] Another object of the present invention is to provide a large-scale unmanned boat scheduling system, which can construct a model drift index according to the error window fluctuation, perform drift judgment based on the model drift index and trigger model self-calibration, and solve the problems of scheduling error accumulation and drift correction lag in the current unmanned boat scheduling system technology.

[0023] As a preferred solution of the large-scale unmanned boat scheduling system described in the present invention, among them: the disturbance prediction model construction module, the control data propulsion module, the disturbance prediction model construction module is used to establish a disturbance prediction model to predict the attitude change of the platform through input parameters such as propulsion angle, power, current water depth, and acceleration, assign different weights to the input factors, adjust the coefficients through historical data learning, perform normalization processing and predict the disturbance according to the stable interval and change of the control parameters, and the model outputs different disturbance prediction thresholds. The control data propulsion module is used to perform non-linear fusion mapping on the propulsion angle, propulsion power, acceleration, and depth, construct a function structure for predicting the disturbance, estimate the possible attitude response of the platform based on dynamic weights, and evaluate the influence of the included angle between the propulsion direction and the heading, the change trend of the acceleration, and the water depth on the platform control reaction; the attitude stripping module includes a stripping residual analysis module and an abnormal disturbance compression module. The stripping residual analysis module is used to compare the actual attitude change of the platform collected by the sensor with the result of the disturbance prediction model, analyze the stripping residual, perform different degrees of weight weakening or suppression according to different residual values, and finally obtain the real platform change after removing the influence of equipment control. The abnormal disturbance compression module is used to adopt different compression methods according to different error intervals based on the preset error level standard, reduce the influence of external interference on the model, execute a decreasing compression rate through an increasing function, and retain the real part of the water body disturbance; the drift judgment module includes a drift index generation module and a drift judgment and self-calibration module. The drift index generation module is used to continuously monitor the change of the error through the sliding time window mechanism, construct a model drift index, update the drift index according to the comparison of the current error data with the previous time period, and dynamically adjust the time span and update frequency of the sliding window. The drift judgment and self-calibration module is used to perform drift judgment based on the model drift index, and adopt different processing methods according to different drift thresholds. After the model is updated, error verification and response testing are carried out to confirm the stability of the prediction error.

[0024] A computer device includes a memory and a processor. The memory stores a computer program, and the execution of the computer program by the processor implements the steps of a large-scale unmanned boat scheduling method.

[0025] A computer-readable storage medium stores a computer program thereon. When the computer program is executed by a processor, it implements the steps of a large-scale unmanned boat scheduling method.

[0026] Advantages of the present invention: The large-scale unmanned boat scheduling method provided by the present invention constructs a disturbance prediction model based on propulsion angle and power parameters, maps propulsion control data using a non-linear characteristic function, provides high-precision disturbance prediction, improves the adaptability of the model under dynamic water conditions, analyzes and strips residuals based on sensor data to obtain the stripped attitude signal, and compresses abnormal disturbances, improving the system's response ability to environmental changes, providing a more stable and reliable unmanned boat scheduling ability in a dynamically changing water environment, constructs a model drift index based on error window fluctuations, performs drift judgment based on the model drift index and triggers model self-calibration, avoiding scheduling errors caused by model drift, and ensuring the long-term stable operation of the unmanned boat scheduling system in complex and changing environments. The present invention achieves better results in terms of disturbance prediction accuracy, error stripping and compression, and model drift correction and self-calibration. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] To more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0028] Figure 1 It is the overall flowchart of the large-scale unmanned boat scheduling method provided by the first embodiment of the present invention.

[0029] Figure 2 It is the overall flowchart of the large-scale unmanned boat scheduling system provided by the third embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0030] To make the above objects, features, and advantages of the present invention more obvious and understandable, the following will provide a detailed description of the specific embodiments of the present invention with reference to the accompanying drawings of the specification. Obviously, the described embodiments are some embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the scope of protection of the present invention.

[0031] Embodiment 1, referring to Figure 1, which is an embodiment of the present invention, provides a large-scale unmanned boat scheduling method, including:

[0032] S1: Construct a disturbance prediction model based on the propulsion angle and power parameters, and use the non-linear characteristic function to map the propulsion control data.

[0033] Furthermore, constructing the disturbance prediction model includes collecting input control parameters such as the unmanned boat propulsion direction angle, propeller power, current water depth, and platform acceleration as input factors for disturbance prediction, and establishing a disturbance response model to predict the attitude change of the platform.

[0034] Adopt a response construction method with non-linear characteristics, assign different weights to various input factors to measure the degree of influence on the platform attitude change, obtain the adjustment coefficient through historical data learning, and normalize the input. Present the disturbance prediction result through the attitude angle change.

[0035] It should be noted that using the non-linear characteristic function to map the propulsion control data includes fusing and mapping the control parameters including the propulsion angle, propulsion power, acceleration, and depth in a non-linear relationship to form a function structure for predicting disturbances, and estimating the possible attitude response of the platform through the dynamic weights of different input parameters.

[0036] The fusion mapping includes judging the possibility of platform deflection by evaluating the angle between the propulsion direction and the platform heading through the system, and judging whether there are sharp turn and sudden acceleration dynamic instability events based on the destructive impact of the propulsion power on the platform stability within a certain range and combining the acceleration change trend.

[0037] Construct the amplification effect of depth on disturbances into a non-linear enhancement structure to simulate the amplification effect of the water depth environment on the platform control reaction.

[0038] Reflect the randomness and uncertainty of disturbances through the fusion of the disturbance response deviation control mechanism. When the disturbance prediction model runs, obtain the real-time disturbance output according to the specific parameter combination, and classify and process it by setting the response interval.

[0039] It should also be noted that a preferred scheme for presenting the disturbance prediction result through the attitude angle change specifically includes when P t <100W, D t <500m, Output When, it means that the platform is sailing in a straight line in low power and shallow water area, the disturbance prediction is very low, and the platform tilt basically comes from environmental factors; when P t >300W, D t ~3000m, When it indicates that the platform is in high power and deep water area and performs a sharp turn operation, the disturbance mainly comes from its own operation and should be removed with emphasis in the subsequent peeling steps; when V t fluctuates violently, resulting in erf(V t ) approaching ±1, it indicates that the platform accelerates abnormally or performs a braking action, which will significantly increase the disturbance output and reflect potential structural resonance or short-term unstable control. Engineering suggestion: Set an upper limit judgment threshold. If , then a large amount of peeling is required; if for a long time , it indicates that there is almost no interference from the device itself, and the peeling intensity should be reduced.

[0040] It should be noted that an optimal scheme for estimating the possible attitude response of the platform by combining the dynamic weights of different input parameters specifically includes calculating the attitude change caused by the predicted disturbance which is expressed as:

[0041]

[0042] Among them, is the predicted attitude disturbance angle of the platform at time t, representing the inclination or rotation change caused by the action of the submersible unmanned vehicle. t is the time index and discrete sampling point. is the propulsion direction angle, representing the angle between the propulsion direction of the unmanned vehicle and the platform coordinate system. P t is the propulsion power, D t is the diving depth, V t is the current acceleration of the platform. is the fitting coefficient of the direction angle factor, which needs to be obtained by training with measured data. α p is the fitting coefficient of the propulsion power factor, used to adjust the influence of the propulsion power on the attitude disturbance. α d is the fitting coefficient of the depth factor, characterizing the non-linear effect of the depth on the disturbance amplification. erf(V t ) is the Gaussian error function, which is expressed as:

[0043]

[0044] Among them, u represents the integration variable, and erf(V t ) is used to model the uncertainty of the disturbance and the amplification effect of the small disturbance.

[0045] It should also be noted that by constructing a disturbance prediction model based on the propulsion angle and power parameters, the system can use various input control parameters such as thruster angle, power, acceleration, and water depth to establish a non-linear disturbance prediction model to predict the platform attitude change; by assigning different weights to these input parameters and combining historical data for learning, the system can accurately predict the attitude change of the platform under different working conditions; when the system evaluates the influence of the included angle between the propulsion direction and the platform heading, the acceleration change trend, and the water depth on the platform, it can fuse these control data into a highly reliable disturbance prediction model based on the non-linear characteristic function; this step, through the application of the non-linear characteristic function, ensures that in various dynamic situations, such as sharp turns, acceleration, etc., the attitude disturbance caused by the platform action can be accurately predicted; this not only improves the stability of the platform but also provides a reliable disturbance prediction basis for subsequent disturbance stripping and drift judgment; through this high-precision model, it can ensure the efficient operation and attitude stability of the unmanned boat in a complex environment, with applicability and accuracy.

[0046] S2: Analyze and strip the residuals based on the sensor data to obtain the stripped attitude signal and compress the abnormal disturbances.

[0047] Furthermore, analyzing and stripping the residuals based on the sensor data includes that the platform collects the actual attitude change angle in real time through the sensor and compares it with the predicted disturbance result provided by the disturbance prediction model to analyze and strip the residuals.

[0048] It should be noted that compressing the abnormal disturbances includes performing segmented compression calculations according to the preset error level standard; combining the compression mechanism and the error, using an increasing function to execute a decreasing compression rate, and finally in the output signal, retaining the real water body disturbance part and actively stripping the high-amplitude error caused by abnormal equipment operation.

[0049] It should also be noted that a preferred scheme for analyzing and stripping the residuals specifically includes calculating the net water body disturbance signal after stripping Expressed as:

[0050]

[0051] Wherein, represents the net water body environment attitude signal after stripping of the unmanned boat floating platform at time t, that is, after removing the equipment control disturbance, the remaining real attitude change caused by the external water body, represents the actual attitude change angle measured by the platform attitude sensor at time t, and ln(·) is the natural logarithm function, which is used to perform non-linear compression processing on the square of the error term, represents the square of the prediction error, which reflects the difference between the platform response and the model prediction. The larger it is, the more it indicates that the model fails to fully capture the disturbance influence.

[0052] It should also be noted that a preferred solution for segmented compression calculation according to the preset error level standard specifically includes when When When, it indicates that the model prediction is accurate, the stripping function has almost no compression, and all environmental signal fluctuations are retained; when the error is between 0.2 - 0.4 rad, it indicates that medium residuals appear, the function compresses, avoiding the residuals from dominating the system judgment, and the environmental signals are retained but leveled to a certain extent; when the error > 1 rad, it indicates that abnormal residuals appear, and the function compression rate is significant, expressed as:

[0053]

[0054] Where, ∈ t represents the instantaneous prediction error value of the disturbance prediction model at time t. When the error is large (such as 1.5), the system soft - limits it through the logarithmic compression function in the denominator, so that the finally output environmental disturbance will not be wrongly amplified due to large errors; the result is approximately 0.83, indicating that although the original error is large, the system controls its influence through the mechanism and outputs a leveled disturbance value to avoid false stripping or misjudgment.

[0055] It should also be noted that by collecting the actual attitude changes of the platform in real - time and comparing with the prediction results of the disturbance prediction model, the system can analyze and remove the disturbances generated by platform control; residual analysis helps the system determine which disturbances come from the device itself and which are caused by external environmental factors. Based on different residual values, the system adopts different processing strategies, thereby accurately removing the influence caused by the device and retaining the external environmental disturbances; this step ensures that the model outputs a signal closer to the actual environmental disturbance rather than the error caused by device anomalies through dynamically stripping disturbances and compressing abnormal disturbances; by reducing the influence of high - amplitude errors through the compression mechanism, the platform's ability to identify real water body disturbances is improved, thereby improving the accuracy and robustness of environmental perception.

[0056] S3: Construct a model drift index according to the error window fluctuation, and perform drift judgment based on the model drift index and trigger model self - calibration.

[0057] Furthermore, constructing the model drift index includes using a sliding time - window mechanism to continuously monitor the change of model errors, and constructing the model drift index by evaluating the cumulative deviation and fluctuation amplitude of errors within the current time period.

[0058] According to the error average value and the difference in error changes before and after, and performing integral processing based on a certain time interval, to capture whether there is a systematic error accumulation phenomenon in the disturbance prediction model. When detecting a new error data point during operation, the new error data point is incorporated into the current sliding window, compared with the previous time period for error, and the drift index is updated.

[0059] Dynamically adjust the time span and update frequency of the sliding window according to the task cycle.

[0060] It should be noted that triggering model self-calibration includes making a drift judgment based on the model drift index and taking different processing methods according to different drift thresholds.

[0061] If the existing disturbance prediction model structure does not adapt to the current change of disturbance characteristics, the training module is called to reload the recently collected control input and platform response data, update the fitting parameters and adjust the model structure, and construct a new round of disturbance prediction model.

[0062] After the disturbance prediction model is updated, error verification and response testing are carried out to confirm whether the prediction error returns to stability. If not up to standard, roll back to the previous version and continue to run.

[0063] It should also be noted that a preferred scheme for constructing the model drift index specifically includes calculating the model drift monitoring index Ω t , expressed as:

[0064]

[0065] where, Ω t is the disturbance prediction model drift index of the unmanned boat platform at time t. The larger the value, the stronger the error instability, and it may be necessary to trigger model self-calibration. ε t is the prediction error of the unmanned boat platform at time t. δ τ represents the change rate of the disturbance error at time τ, that is, the difference between the previous and subsequent error values, reflecting the fluctuation intensity of the prediction error. τ is the time index within the integration interval, used to traverse the error change from t - T w to t. dτ is the variable of time integration, indicating the cumulative processing of the error within the time period. By integrating the error term, the disturbance error amount within the entire time window is calculated. k is the historical maximum error time offset, and T w represents the length of the sliding time window, which is the time period referred to by the unmanned boat system to detect the error fluctuation trend. η is a small positive value to prevent division by zero. If Ω t > ξ, the system automatically triggers self-calibration or retraining.

[0066] It should also be noted that a preferred scheme for making a drift judgment based on the model drift index and taking different processing methods according to different drift thresholds specifically includes when Ω t ≤0.5, it indicates that the system error value is small and the fluctuation is also small. The model is stable and the error is controllable. The current model can be continued to be used without intervention; when 0.5 < Ω tWhen Ω ≤ 1.0, the maximum error has a certain value, but the error fluctuation is still within an acceptable range, indicating a slight drift. Record the warning flag and manually review the model status if necessary; when t Ω > 1.0, it indicates that the error may be mutating, fluctuating violently, or deviating for a long time, which may be caused by system hardware drift, environmental mutation, algorithm inadaptability, etc. Automatically trigger the model retraining or rollback mechanism, replace it with the previous stable model or retrain a new model based on the current data.

[0067] It should also be noted that during the process of constructing the model drift index from the error window fluctuation, the system continuously monitors the change of the model prediction error through a sliding time window, calculates the drift index, and judges whether there is an error accumulation phenomenon in the model according to this index; if the drift index is greater than the preset threshold, the system will trigger the self-correction process of the model to ensure the continuous accuracy of the model; the drift judgment mechanism can take different adjustment measures according to different drift situations, such as automatically updating the model or rolling back to a more stable version, to ensure the stability and efficiency of the unmanned boat; this step can effectively detect and correct the drift problem that may occur in the long-term operation of the model. Through the combination of the sliding time window and the drift index, the system can adjust the model in real time to avoid the cumulative impact of errors on the prediction results; the introduction of the self-calibration mechanism ensures that the system can still maintain high-precision disturbance prediction under the conditions of equipment aging, environmental change or algorithm failure, and ensures the long-term stability and reliability of the system, improving the ability of the unmanned boat to operate autonomously in complex marine environments for a long time.

[0068] Embodiment 2 is an embodiment of the present invention, which provides a large-scale unmanned boat scheduling method. In order to verify the beneficial effects of the present invention, scientific demonstrations are carried out through economic benefit calculation and simulation experiments.

[0069] First, an integrated test platform including a control input module, an environment simulation system, and an attitude feedback mechanism was experimentally constructed. A total of seven groups of typical operation scenarios were set up in the experiment, covering various types of conditions such as low-power straight navigation to high-power sharp turns in deep water operations, ensuring that the adaptability and response accuracy of the model under different input combinations could be effectively tested. Before the experiment started, the unmanned boat system was first connected to the fully simulated environmental chamber, which could adjust simulation environmental variables such as water depth, resistance parameters, and disturbance fluctuation frequencies. The platform was equipped with a high-precision propulsion angle adjustment system and a variable-frequency propulsion power control unit to achieve combined inputs of different angles, powers, and depths. Each group of tests was input according to preset parameters such as propulsion angle, propulsion power, underwater depth, and acceleration changes, and the original values of all variables were recorded in real time in the system background. Subsequently, through the disturbance prediction system, the attitude change trend of the unmanned boat under different control parameters was modeled and predicted. This process not only considered the influence of the angle between the propeller direction and the main axis of the boat body on attitude stability, but also comprehensively evaluated the influence degree of propulsion power on hydrodynamic response, as well as the control delay and amplification effect caused by changes in underwater depth. The platform also continuously monitored the acceleration fluctuations of the unmanned boat during propulsion to determine whether there were unstable events such as sharp turns, sudden accelerations, or emergency decelerations. During the process of each group of experiments, the platform attitude sensor synchronously collected the actual attitude angle changes and compared them with the results calculated by the disturbance prediction model. Once a large prediction error was detected, the stripping module would be automatically activated. Combining with the built-in compression mechanism, it filtered out the high-amplitude errors caused by abnormal device control operations, thereby separating the disturbance signals closer to those caused by the real water environment. The compression mechanism was designed as a segmented response method, applying different degrees of dynamic compression to different error levels to ensure that the disturbance information output by the system could balance accuracy and stability. All tests were carried out under the same simulated water temperature and water flow density conditions, ensuring the consistency and comparability of the data. At the same time, the system exported and recorded all key indicators and processing results of each test, providing a solid basis for subsequent data analysis and comparison. Through this experimental process, the response characteristics of the disturbance prediction model in multi-input and multi-interference scenarios could be systematically evaluated, and the improvement effect of the present invention on the stable control of the unmanned boat platform in complex environments could be verified. Referring to Table 1, some experimental data were recorded and analyzed.

[0070] Table 1 Experimental Data Record Sheet

[0071]

[0072] As can be seen from the table, the non-linear disturbance prediction model constructed by fusing multiple input variables such as propulsion angle, power, acceleration, and water depth can accurately reflect the attitude change trend of the unmanned boat platform under different operating environments. The experimental table shows the differences between the predicted disturbances of the system and the actual measured values under different test conditions, as well as the net water body disturbance signals after being processed by the stripping mechanism; in the case of stable navigation in low power and shallow water areas (such as experimental numbers T001 to T003), the error between the predicted disturbance output by the system and the actual measured value is very small, and the maximum error does not exceed 0.05 rad. The stripped net disturbance signal is almost the same as the predicted value; this indicates that in such a simple and stable operating environment, the prediction model shows extremely high accuracy; the experimental data shows that the disturbance prediction error is efficiently controlled within a low range by the system, and the stripping mechanism hardly requires further compression processing, and the system can retain all effective environmental signals; however, when performing sharp turning operations in high power deep water areas (such as experimental numbers T004 to T007), with the increase of propulsion power and the deepening of water depth, the attitude change of the platform becomes more complex; especially when the platform performs operations such as large-scale acceleration or sharp turning, the predicted disturbance of the system increases rapidly, and the maximum predicted value reaches 1.1 rad; at this time, the system can effectively suppress the high-amplitude error caused by abnormal equipment control through the error stripping mechanism, so as to only retain the real changes caused by external water body disturbances. For example, in experimental number T007, although the actual measured value is 1.25 rad and the predicted disturbance value is 1.1 rad, after being processed by the system, the stripped net disturbance value is only 0.83 rad, which is significantly lower than the original predicted value; this indicates that the system has successfully filtered the error caused by unstable equipment dynamics through the compression mechanism, ensuring more accurate and stable identification of attitude disturbances.

[0073] Compared with the prior art that only predicts disturbances based on linear power, the present invention introduces the propulsion direction angle, depth enhancement mechanism, and non-linear acceleration response, improving the multi-dimensional fusion ability of attitude disturbance modeling and having strong innovation; in terms of data compression, different from the traditional mean smoothing method, the present invention uses adaptive compression mapping to dynamically adjust the compression amplitude based on the error level, significantly reducing the risk of false stripping while retaining effective information; in addition, the upper and lower limit response intervals set in the experiment enable the model to intelligently classify and identify states such as "normal disturbance", "medium deviation", and "abnormal dynamics", and output decision signals accordingly, providing accurate basis for subsequent automatic navigation and stripping algorithms. In summary, the present invention fully reflects the technical innovation and system robustness of the disturbance prediction and stripping model, verifying its wide applicability and promotion value in complex unmanned boat operating environments.

[0074] Example 3, refer to Figure 2, which is an embodiment of the present invention, provides a large-scale unmanned boat scheduling system, a disturbance prediction and data propulsion module 100, an attitude stripping module 200, and a drift judgment module 300.

[0075] Among them, S4: The disturbance prediction and data propulsion module 100 includes a disturbance prediction model construction module 101 and a control data propulsion module 102. The disturbance prediction model construction module 101 is used to establish a disturbance prediction model to predict the attitude change of the platform through input parameters such as propulsion angle, power, current water depth, and acceleration, assign different weights to the input factors, adjust the coefficients through historical data learning, perform normalization processing, and predict the disturbance. According to the stable interval and change situation of the control parameters, the model outputs different disturbance prediction thresholds. The control data propulsion module 102 is used to perform non-linear fusion mapping on the propulsion angle, propulsion power, acceleration, and depth, construct a function structure for predicting the disturbance, dynamically estimate the possible attitude response of the platform, and evaluate the influence of the included angle between the propulsion direction and the heading, the acceleration change trend, and the water depth on the platform control reaction.

[0076] It should be noted that the disturbance prediction model construction module 101 is responsible for establishing a disturbance model for predicting the attitude change of the platform by collecting the control data of the unmanned boat (including input parameters such as propulsion angle, power, current water depth, and acceleration); by assigning different weights to each input factor, the model adjusts these weight coefficients through historical data learning, so as to more accurately predict the disturbance of the platform under different working conditions. The system will output the threshold of the disturbance prediction value according to the stable interval and change situation of the control parameters to help identify whether the system is in an abnormal operation state; the control data propulsion module 102 performs non-linear fusion mapping on the input control data (propulsion angle, power, acceleration, depth). Through this step, the model dynamically estimates the possible attitude response of the platform, evaluates the influence of the included angle between the propulsion direction and the heading, the acceleration change trend, and the water depth on the platform control reaction. The disturbance prediction and data propulsion module 100 ensures that the system can predict and feedback possible disturbances in real time in various complex working environments by synthesizing a function structure for predicting the disturbance, so that the platform can always maintain stability and efficiency when performing tasks.

[0077] The disturbance prediction and data propulsion module 100 collects control data such as propulsion angle, power, acceleration, and water depth to establish a model for predicting the attitude change of the platform and outputs the disturbance prediction value; the attitude stripping module 200 uses these prediction values, calculates the residual by comparing the actual measurement data with the prediction data, and strips the disturbance caused by the platform control, retaining the real water body disturbance signal; through this linkage, the system can accurately distinguish the operation disturbance of the platform and the disturbance of the external environment, providing accurate data for subsequent control and environmental perception.

[0078] S5: The attitude stripping module 200 includes a stripping residual analysis module 201 and an abnormal disturbance compression module 202. The stripping residual analysis module 201 is used to analyze the stripping residuals by comparing the actual attitude changes of the platform collected by the sensor with the results of the disturbance prediction model, and perform different degrees of weight weakening or suppression according to different residual values, and finally obtain the real platform changes after removing the influence of device control. The abnormal disturbance compression module 202 is used to reduce the influence of external interference on the model by adopting different compression methods according to different error intervals based on the preset error level standard, and execute a decreasing compression rate through an increasing function to retain the real water body disturbance part.

[0079] It should be noted that the function of the attitude stripping module 200 is to remove the influence of the platform control system itself on the attitude changes and retain the real disturbance signals caused by external water factors, so as to obtain more accurate marine environment perception. It includes two key parts: the stripping residual analysis module 201 and the abnormal disturbance compression module 202. The stripping residual analysis module 201 first collects the actual attitude change data of the platform in real time through the sensor and compares it with the output results of the disturbance prediction model. Through comparative analysis, the system can obtain the residual values, which indicate the influence degree of the platform control system. According to the size of the residuals, the system performs different degrees of weight weakening or suppression on different residuals, so as to obtain the real platform attitude changes after removing the influence of device control. This process ensures that the system can accurately distinguish external water body disturbances from platform self-operation disturbances, and thus make correct responses to marine environment changes; on this basis, the abnormal disturbance compression module 202 is responsible for compressing the abnormal disturbances. The system applies different compression methods to different error intervals according to the preset error level standard. When the error is small, the system retains more original data, and when the error is large, an increasing function is used to gradually reduce the compression rate to reduce the influence of external interference on the model. Finally, the system can retain the real water body disturbance signal and eliminate the high-amplitude errors caused by abnormal device operations.

[0080] The attitude stripping module 200 removes the disturbances of the platform control and removes the high-amplitude errors caused by device abnormalities through the abnormal disturbance compression module, leaving the real water body disturbances; the drift judgment module 300 relies on the stripped net disturbance signal, constructs a drift index by monitoring the error changes, and judges whether the system drifts. If the drift index exceeds the threshold, the system will trigger a self-calibration or rollback mechanism to restore the stability and prediction accuracy of the model.

[0081] S6: The drift judgment module 300 includes a drift index generation module 301 and a drift judgment and self-calibration module 302. The drift index generation module 301 is used to continuously monitor the change of errors through a sliding time window mechanism, construct a model drift index, update the drift index based on the comparison between the current error data and that of the previous time period, and dynamically adjust the time span and update frequency of the sliding window. The drift judgment and self-calibration module 302 is used to perform drift judgment based on the model drift index, and adopt different processing methods according to different drift thresholds. After the model is updated, error verification and response testing are carried out to confirm the stability of the prediction error.

[0082] It should be noted that the role of the drift judgment module 300 is to detect whether the disturbance prediction model drifts and perform self-calibration when necessary; the drift index generation module 301 continuously monitors the change of errors through a sliding time window mechanism, generates a drift index by comparing the current error data with the error data of the previous time period, and dynamically adjusts the time span and update frequency of the sliding window according to the change of the drift index; in this way, the system can track the fluctuation of the model error in real time, update the drift index according to the actual situation, and maintain high sensitivity to the disturbance prediction model; when the drift index reaches a certain threshold, the drift judgment and self-calibration module 302 will be activated, and it will decide whether to perform self-calibration of the model based on the result of the drift judgment; different processing methods are adopted according to the drift threshold: if the drift is small, the system will record an alarm flag and conduct manual review; if the drift is large, the system will automatically trigger the re-training or rollback mechanism of the model, and ensure the stable prediction accuracy of the model by replacing it with the previous stable model or re-training a new model; after the model is updated, the system will conduct error verification and response testing to confirm whether the model has returned to a stable state and ensure the reliability of subsequent prediction results.

[0083] The disturbance prediction model construction module 100 is responsible for outputting the prediction results of the attitude disturbances that may occur on the platform, providing data support for the drift judgment module. The drift judgment module 300 calculates the drift index by comparing the historical error with the current error to judge whether the model drifts; when the drift index is too high, the module will trigger the model self-calibration or rollback mechanism to ensure the accuracy and stability of the model during long-term operation.

[0084] If a function is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods of various embodiments of the present invention. The aforementioned storage medium includes: USB flash drives, mobile hard disks, read-only memories (ROM, Read-Only Memory), random access memories (RAM, Random Access Memory), magnetic disks, or optical discs, etc., various media that can store program codes.

[0085] The logic and / or steps represented in the flowchart or described in other ways herein, for example, can be considered as a predefined sequence list of executable instructions for implementing logical functions. It can be specifically implemented in any computer-readable medium for use by an instruction execution system, apparatus, or device (such as a computer-based system, a system including a processor, or other systems that can fetch instructions from the instruction execution system, apparatus, or device and execute the instructions), or used in combination with these instruction execution systems, apparatuses, or devices. For the purposes of this specification, a "computer-readable medium" can be any device that can contain, store, communicate, propagate, or transmit a program for use by or in combination with an instruction execution system, apparatus, or device.

[0086] More specific examples (non-exhaustive list) of computer-readable media include the following: electrical connection parts with one or more wirings (electronic devices), portable computer disk cartridges (magnetic devices), random access memories (RAM), read-only memories (ROM), erasable programmable read-only memories (EPROM or flash memories), optical fiber devices, and portable compact disc read-only memories (CDROM). Additionally, a computer-readable medium can even be paper or other suitable media on which a program can be printed, because the program can be obtained electronically, for example, by optically scanning the paper or other media, then editing, interpreting, or otherwise processing it as appropriate, and then storing it in a computer memory.

[0087] It should be understood that each part of the present invention can be implemented by hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, any one or a combination of the following techniques well known in the art can be used: discrete logic circuits having logic gate circuits for implementing logical functions on data signals, application specific integrated circuits having appropriate combinational logic gate circuits, programmable gate arrays (PGAs), field programmable gate arrays (FPGAs), etc. It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present invention, and they should all be covered by the scope of the claims of the present invention.

Claims

1. A method for scheduling large-scale unmanned boats, characterized in that including: Construct a disturbance prediction model based on the propulsion angle and power parameters, and use a non-linear characteristic function to map the propulsion control data; Analyze and strip the residuals according to the sensor data to obtain the stripped attitude signal and compress abnormal disturbances; Construct a model drift index based on the error window fluctuation, and perform drift judgment and trigger model self-calibration based on the model drift index; Performing drift judgment includes continuously monitoring the change of the model error, dynamically evaluating the error fluctuation, analyzing the error in the current time period, accumulating the deviation and calculating the fluctuation amplitude, using the average value and the difference of the front and back changes of the error to construct a disturbance prediction model drift index to help judge whether there is a systematic error in the disturbance prediction model, and dynamically adjusting the time span and update frequency of the sliding window according to the task cycle; Triggering model self-calibration includes performing drift judgment based on the model drift index, taking different processing methods according to different drift thresholds, performing error verification and response tests after the model is updated to determine that the updated model runs stably, and rolling back to the previous version if the standard is not met.

2. The large-scale unmanned boat scheduling method according to claim 1, wherein: The constructing of the disturbance prediction model includes, By collecting the input control parameters of the unmanned boat propulsion direction angle, propeller power, current water depth and platform acceleration as the input factors for disturbance prediction, and establishing a disturbance response model to predict the attitude change of the platform; Adopt a response construction method with non-linear characteristics, assign different weights to various input factors to measure the influence degree on the platform attitude change, obtain the adjustment coefficient through historical data learning method, perform normalization processing on the input, and present the disturbance prediction result through the attitude angle change.

3. The method for scheduling large-scale unmanned boats according to claim 1 or 2, characterized in that: The using of the non-linear characteristic function to map the propulsion control data includes, Fuse and map the control parameters including propulsion angle, propulsion power, acceleration and depth in a non-linear relationship to form a function structure for predicting disturbances, and estimate the possible attitude response of the platform through the dynamic weights of different input parameters; The fusion mapping includes judging the possibility of platform deflection by system evaluation of the angle between the propulsion direction and the platform heading, based on the destructive influence of the propulsion power on the platform stability within a certain range, and combining the acceleration change trend to judge whether there are sharp turns, sudden acceleration and dynamic instability events; Construct the amplification effect of depth on disturbance into a non-linear enhancement structure to simulate the amplification effect of water depth environment on platform control reaction; Reflect the randomness and uncertainty of disturbances through a fusion disturbance response deviation control mechanism. When the disturbance prediction model runs, obtain the real-time disturbance output according to the specific parameter combination, and perform classification processing by setting the response interval.

4. The large-scale unmanned boat scheduling method according to claim 3, wherein: The analyzing and stripping of residuals according to the sensor data includes, The platform collects the actual attitude change angle in real time through the sensor, compares it with the predicted disturbance result provided by the disturbance prediction model, and analyzes and strips the residuals.

5. The large-scale unmanned boat scheduling method according to any one of claims 1, 2 or 4, characterized in that: The compressing of abnormal disturbances includes, Perform segmented compression calculation according to the preset error level standard; Combine the compression mechanism with the error, use an increasing function to execute a decreasing compression rate. In the final output signal, retain the real water body disturbance part and actively strip the high-amplitude error generated by abnormal equipment operation.

6. The large-scale unmanned boat scheduling method according to claim 5, wherein: The constructing of the model drift index includes, Adopt a sliding time window mechanism to continuously monitor the change of model error. By evaluating the cumulative deviation and fluctuation amplitude of the error in the current time period, construct a model drift index; According to the average error and the difference in error change before and after, and perform integral processing based on a certain time interval to capture whether there is a systematic error accumulation phenomenon in the disturbance prediction model. When a new error data point is detected during operation, incorporate the new error data point into the current sliding window, compare the error with the previous time period, and update the drift index; Dynamically adjust the time span and update frequency of the sliding window according to the task cycle.

7. The large-scale unmanned boat scheduling method according to any one of claims 1, 2, 4 or 6, characterized in that: The trigger for model self-calibration includes, Perform drift judgment based on the model drift index and adopt different processing methods according to different drift thresholds; If the existing disturbance prediction model structure does not adapt to the current change of disturbance characteristics, reload the recently collected control input and platform response data, update the fitting parameters and adjust the model structure, and construct a new round of disturbance prediction models; After the disturbance prediction model is updated, perform error verification and response testing to confirm whether the prediction error returns to stability. If not up to standard, roll back to the previous version and continue to run.

8. Large-scale unmanned boat scheduling system, characterized in that: Include a disturbance prediction and data advancement module (100), an attitude stripping module (200), and a drift judgment module (300); The disturbance prediction and data advancement module (100) includes a disturbance prediction model construction module (101) and a control data advancement module (102). The disturbance prediction model construction module (101) is used to establish a disturbance prediction model to predict the attitude change of the platform through input parameters such as propulsion angle, power, current water depth, and acceleration, assign different weights to the input factors, adjust the coefficients through learning from historical data, perform normalization processing and predict the disturbance. According to the stable interval and change of the control parameters, the model outputs different disturbance prediction thresholds. The control data advancement module (102) is used to perform non-linear fusion mapping on the propulsion angle, propulsion power, acceleration, and depth, construct a function structure for predicting the disturbance, estimate the possible attitude response of the platform based on dynamic weights, and evaluate the influence of the included angle between the propulsion direction and the heading, the change trend of acceleration, and the water depth on the platform control reaction; The attitude stripping module (200) includes a stripping residual analysis module (201) and an abnormal disturbance compression module (202). The stripping residual analysis module (201) is used to compare the actual attitude change of the platform collected by the sensor with the result of the disturbance prediction model, analyze the stripping residual, and perform different degrees of weight weakening or suppression according to different residual values, and finally obtain the true change of the platform after removing the influence of equipment control. The abnormal disturbance compression module (202) is used to adopt different compression methods according to different error intervals based on the preset error level standard, reduce the influence of external interference on the model, execute a decreasing compression rate through an increasing function, and retain the true part of the water body disturbance; The drift judgment module (300) includes a drift index generation module (301) and a drift judgment and self-calibration module (302). The drift index generation module (301) is used to continuously monitor the change of errors through a sliding time window mechanism, construct a model drift index, update the drift index based on the comparison between the current error data and the previous time period, and dynamically adjust the time span and update frequency of the sliding window. The drift judgment and self-calibration module (302) is used to perform drift judgment based on the model drift index, and adopt different processing methods according to different drift thresholds. After the model is updated, error verification and response testing are carried out to confirm that the prediction error is stable.

9. A computer device, comprising a memory and a processor, the memory storing a computer program, characterized in that, When the processor executes the computer program, the steps of the large-scale unmanned boat scheduling method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, the steps of the large-scale unmanned boat scheduling method according to any one of claims 1 to 7 are implemented.

Images