Multi-sensor fusion method and system based on shipborne six-degree-of-freedom stable platform

Abstract

The present application discloses a multi-sensor fusion method and system based on a shipborne six-degree-of-freedom stabilized platform, the method comprising: acquiring environmental data and platform status data; calculating a first information entropy of the environmental data and a second information entropy of the platform status data; calculating a consistency index between the environmental data and the platform status data; determining a probability density of pseudo data according to the first information entropy, the second information entropy and the consistency index; if the probability density is greater than a preset threshold, performing Gaussian distribution fusion on the environmental data and the platform status data to obtain a first environmental perception feature group; if the probability density is less than or equal to the preset threshold, performing centralized fusion on the environmental data and the platform status data to obtain a second environmental perception feature group; determining a future motion trend of the shipborne six-degree-of-freedom stabilized platform through the first environmental perception feature group or the second environmental perception feature group; and performing balance compensation on the shipborne six-degree-of-freedom stabilized platform according to the future motion trend.

Description

Technical Field

[0001] The embodiments of the present application relate to the field of data fusion, and in particular to a multi-sensor fusion method and system based on a shipborne six-degree-of-freedom stabilized platform. Background Art

[0002] When a ship is sailing or operating at sea, it will inevitably produce six degrees of freedom movements, namely roll, pitch, bow pitch, sway, swing and heave, under the action of sea breeze and waves, which will affect the activities on the ship, such as the operation of precision equipment on the ship or the lifting of cargo equipment. The shipborne six-degree-of-freedom stabilized platform can provide a stable platform for the activities on the ship, so the shipborne six-degree-of-freedom stabilized platform is widely used on ships.

[0003] The shipborne six-degree-of-freedom stabilized platform can compensate and isolate the motion of the carrier in real time through the servo motion control system to ensure the relative stability of the platform. However, in complex marine environments, the servo motion control system may lead to poor compensation when faced with unreliable or disturbed sensor data. For example, noise in the environment or sensor failure may generate false data, which will cause errors in the stability control of the shipborne six-degree-of-freedom stabilized platform and affect normal activities on the ship.

[0004] There is currently no better solution to the above problems. Summary of the invention

[0005] The embodiments of the present application provide a multi-sensor fusion method and system based on a shipborne six-degree-of-freedom stabilized platform, which are used to effectively improve the accuracy of stability control of the shipborne six-degree-of-freedom stabilized platform and reduce the impact on normal activities on the ship.

[0006] To achieve the above objectives, the embodiments of the present application adopt the following technical solutions:

[0007] In a first aspect, a multi-sensor fusion method based on a shipborne six-degree-of-freedom stabilized platform is provided, the method comprising:

[0008] Acquiring multi-sensor data through a multi-sensor disposed on a shipborne six-degree-of-freedom stable platform, wherein the multi-sensor data includes environmental data and platform status data;

[0009] respectively calculating a first information entropy of the environment data and a second information entropy of the platform state data to quantify the uncertainty of the environment data and the platform state data;

[0010] Calculating a consistency index between each of the environment data and each of the platform status data to quantify the correlation between the environment data and the platform status data;

[0011] Determining a probability density of pseudo data in the multi-sensor data according to the first information entropy, the second information entropy and the consistency index;

[0012] When the probability density is greater than a preset threshold, performing Gaussian distribution fusion on the environment data and the platform state data to obtain a first environment perception feature group;

[0013] When the probability density is less than or equal to a preset threshold, the environment data and the platform status data are centrally integrated to obtain a second environment perception feature group;

[0014] determining a future motion trend of the shipborne six-degree-of-freedom stabilized platform by using the first environment perception feature group or the second environment perception feature group;

[0015] Performing balance compensation on the shipborne six-degree-of-freedom stable platform according to the future motion trend;

[0016] Among them, the first environmental perception feature group or the second environmental perception feature group includes multiple environmental features, multiple platform state features and multiple potential functions, the potential function is used to connect the environmental features and the platform state features, and the strength of the potential function is determined based on the degree of influence of the environmental features on the platform state features.

[0017] In a possible implementation manner of the first aspect, determining the probability density of pseudo data in the multi-sensor data according to the first information entropy, the second information entropy, and the consistency index includes:

[0018] Calculating, according to the consistency index, an inconsistency index of the environment data and the platform status data, wherein the sum of the inconsistency index and the consistency index is 1;

[0019] Calculating a possibility index of pseudo data according to the first information entropy and the inconsistency index;

[0020] The possibility index is normalized according to the first information entropy, the second information entropy and the consistency index to obtain the probability density of the pseudo data in the multi-sensor data.

[0021] In another possible implementation manner of the first aspect, when the probability density is greater than a preset threshold, performing Gaussian distribution fusion on the environment data and the platform state data to obtain a first environment perception feature group includes:

[0022] When the probability density is greater than a preset threshold, feature extraction is performed on the environment data to obtain a plurality of the environment features E, and feature extraction is performed on the platform state data to obtain a plurality of the platform state features P;

[0023] The environment feature E and the platform state feature P are fused using a Gaussian mixture model to obtain a fused feature group, wherein the fused feature group includes the environment feature Ei and the platform state feature Pj;

[0024] The environmental feature Ei and the platform state feature Pj in the fused feature group are connected through a potential function to obtain the first environmental perception feature group.

[0025] In another possible implementation manner of the first aspect, the step of constructing the Gaussian mixture model includes:

[0026] Determine K Gaussian components, and initialize the model parameters of the Gaussian mixture model for each of the Gaussian components;

[0027] The final model parameters of the Gaussian mixture model are determined by using an expectation maximization algorithm, wherein the final model parameters include a mixture weight, a mean vector, and a covariance matrix;

[0028] Determine a probability density function of each of the Gaussian components according to the mixing weight, the mean vector and the covariance matrix;

[0029] Performing weighted summation on the probability density function of each Gaussian component to determine the overall probability density function of the Gaussian mixture model;

[0030] A Gaussian mixture model is constructed according to the final model parameters and the overall probability density function.

[0031] In another possible implementation manner of the first aspect, when the probability density is less than or equal to a preset threshold, the environment data and the platform state data are centrally integrated to obtain a second environment perception feature group, including:

[0032] When the probability density is less than or equal to a preset threshold, feature extraction is performed on the environment data to obtain a plurality of the environment features E, and feature extraction is performed on the platform state data to obtain a plurality of the platform state features P;

[0033] According to the consistency index, the plurality of environmental features E and the plurality of platform state features P are divided into at least two feature groups, each of the feature groups including the environmental features E and the platform state features P within a preset consistency index range;

[0034] Performing feature fusion on the environment feature E and the platform state feature P in each feature group to obtain a fused feature vector f(E, P);

[0035] Each of the feature vectors f(E, P) is combined with the corresponding potential function to obtain a second environment perception feature group.

[0036] In another possible implementation manner of the first aspect, the performing feature fusion on the environment feature E and the platform state feature P in each feature group to obtain a fused feature vector f(E, P) includes:

[0037] Normalizing the environment feature E and the platform state feature P so that the value ranges of the environment feature E and the platform state feature P are consistent;

[0038] The normalized environmental feature E and the platform state feature P are input into a pre-trained multi-layer perceptron neural network to obtain a fused feature vector f(E, P), wherein the multi-layer perceptron neural network includes an input layer, a hidden layer and an output layer.

[0039] In another possible implementation manner of the first aspect, determining the future movement trend of the shipborne six-degree-of-freedom stabilized platform by using the first environment perception feature group or the second environment perception feature group includes:

[0040] The first environment perception feature group or the second environment perception feature group is input into a preset six-degree-of-freedom rigid body dynamics equation, and the six-degree-of-freedom rigid body dynamics equation is solved by a numerical integration method to obtain the future position, attitude angle, velocity and acceleration of the shipborne six-degree-of-freedom stable platform, wherein the future position, attitude angle, velocity and acceleration are used to describe the future motion trend.

[0041] In another possible implementation manner of the first aspect, performing balance compensation on the shipborne six-degree-of-freedom stable platform according to the future motion trend includes:

[0042] Calculating a position error based on the future position and the acquired target position;

[0043] Calculating a speed error based on the speed and the acquired target speed;

[0044] Calculating an acceleration error according to the acceleration and the acquired target acceleration;

[0045] Calculating an attitude angle error according to the attitude angle and the acquired target attitude angle;

[0046] A control signal is generated according to the position error, the velocity error, the acceleration error and the attitude angle error, so as to perform balance compensation on the shipborne six-degree-of-freedom stable platform through the control signal.

[0047] In a second aspect, the present application provides a controller, comprising:

[0048] a memory configured to store instructions; and

[0049] The processor is configured to call the instructions from the memory and implement the multi-sensor fusion method based on the shipborne six-degree-of-freedom stable platform when executing the instructions.

[0050] In a third aspect, the present application provides a multi-sensor fusion system based on a shipborne six-degree-of-freedom stabilized platform, comprising:

[0051] Controller;

[0052] A shipborne six-degree-of-freedom stabilized platform is connected to the controller.

[0053] Through the above technical scheme, by calculating the information entropy and consistency index of environmental data and platform status data, the probability density of pseudo data can be identified and evaluated. When the probability density of pseudo data is high, the data is processed by Gaussian distribution fusion, thereby reducing the negative impact of pseudo data on the system, improving the accuracy of stability control of the shipborne six-degree-of-freedom stable platform, and reducing the impact on normal activities on the ship. In addition, by adopting two different fusion strategies, Gaussian distribution fusion and centralized fusion, the most suitable data fusion method can be adaptively selected according to the uncertainty and correlation of the data, thereby improving the reliability and accuracy of data fusion, and by processing the fused environmental data and platform status data, the future movement trend of the platform can be more accurately predicted, not only considering the current state of the platform, but also integrating the potential impact of environmental factors on the platform movement, so that effective balance compensation measures can be taken in advance to reduce the occurrence of unstable movement, and the potential function is introduced to connect the environmental characteristics and the platform state characteristics. The strength of the potential function is determined according to the degree of influence of the environmental characteristics on the platform state characteristics, which can model the impact of the environment on the platform state in a more detailed manner, provide a deeper level of feature association, and enhance the response capability of the controller in a complex environment.

[0054] Other features and advantages of the embodiments of the present application will be described in detail in the subsequent specific implementation section. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] Figure 1 A schematic flow chart of a multi-sensor fusion method based on a shipborne six-degree-of-freedom stable platform provided in an embodiment of the present application;

[0056] Figure 2 A schematic diagram of the structure of a multi-sensor fusion system based on a shipborne six-degree-of-freedom stabilized platform provided in an embodiment of the present application. DETAILED DESCRIPTION

[0057] In order to make the purpose, technical scheme and advantages of the embodiments of the present application clearer, the technical scheme in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. It should be understood that the specific implementation methods described herein are only used to illustrate and explain the embodiments of the present application, and are not used to limit the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.

[0058] It should be noted that if the embodiments of the present application involve directional indications (such as up, down, left, right, front, back...), the directional indications are only used to explain the relative position relationship, movement status, etc. between the components under a certain specific posture (as shown in the accompanying drawings). If the specific posture changes, the directional indication will also change accordingly.

[0059] In addition, if there are descriptions involving "first", "second", etc. in the embodiments of the present application, the descriptions of "first", "second", etc. are only used for descriptive purposes and cannot be understood as indicating or suggesting their relative importance or implicitly indicating the number of technical features indicated. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features. In addition, the technical solutions between the various embodiments can be combined with each other, but they must be based on the ability of ordinary technicians in the field to implement them. When the combination of technical solutions is contradictory or cannot be implemented, it should be deemed that such combination of technical solutions does not exist and is not within the scope of protection required by this application.

[0060] Figure 1 The flowchart of a multi-sensor fusion method based on a shipborne six-degree-of-freedom stable platform according to an embodiment of the present application is schematically shown. Figure 1 As shown, an embodiment of the present application provides a multi-sensor fusion method based on a shipborne six-degree-of-freedom stable platform, which is applied to a controller. The method may include the following steps.

[0061] S110, acquiring multi-sensor data through a multi-sensor disposed on a shipborne six-degree-of-freedom stable platform, where the multi-sensor data includes environmental data and platform status data;

[0062] S120, respectively calculating a first information entropy of the environment data and a second information entropy of the platform state data to quantify the uncertainty of the environment data and the platform state data;

[0063] S130, calculating a consistency index between each piece of environmental data and each piece of platform status data to quantify the correlation between the environmental data and the platform status data;

[0064] S140, determining the probability density of pseudo data in the multi-sensor data according to the first information entropy, the second information entropy and the consistency index;

[0065] S150, when the probability density is greater than a preset threshold, performing Gaussian distribution fusion on the environment data and the platform status data to obtain a first environment perception feature group;

[0066] S160: When the probability density is less than or equal to a preset threshold, the environment data and the platform status data are centrally integrated to obtain a second environment perception feature group;

[0067] S170, determining a future movement trend of the shipborne six-degree-of-freedom stabilized platform through the first environment perception feature group or the second environment perception feature group;

[0068] S180, balance compensation of the shipborne six-degree-of-freedom stable platform according to future movement trends;

[0069] Among them, the first environmental perception feature group or the second environmental perception feature group includes multiple environmental features, multiple platform state features and multiple potential functions. The potential function is used to connect the environmental features and the platform state features. The strength of the potential function is determined based on the degree of influence of the environmental features on the platform state features.

[0070] In this embodiment, a variety of sensors are deployed on the shipborne six-degree-of-freedom stabilized platform, including accelerometers, gyroscopes, GPS, laser rangefinders, etc., to detect platform status data of the shipborne six-degree-of-freedom stabilized platform (hereinafter referred to as the platform). For example, the acceleration of the platform can be detected by the accelerometer, the angular velocity change and angle change can be detected by the gyroscope, the position of the platform can be detected by the GPS, and the distance between the platform and the target object can be detected by the laser rangefinder.

[0071] In addition, the platform is also deployed with sensors for detecting environmental data, such as wind speed and direction sensors, wave height sensors, temperature sensors, etc. Among them, the wave height sensor can be a radar or ultrasonic sensor to measure the height of the waves by detecting changes in the height of the water surface.

[0072] This embodiment evaluates the degree of confusion (uncertainty) of environmental data and platform status data by calculating information entropy. Information entropy reflects the magnitude of uncertainty in the data set. A higher information entropy may indicate that there is a lot of noise or false data in the data, and a lower information entropy indicates that the data is more consistent and reliable. Under normal working conditions, the entropy value of sensor data should be kept within a certain range. If the information entropy value increases significantly over a period of time, it may mean that there is false data or data anomalies.

[0073] In addition, since the environment data affects the platform status data, that is, the environment data and the platform status data are related, the correlation between the environment data and the platform status data can be quantified by calculating the consistency index between each environment data and each platform status data.

[0074] Before the specific implementation of S140, it is necessary to perform data preprocessing on the environmental data and the platform status data. The data preprocessing includes preliminary denoising, filtering, etc., so as to preliminarily reduce the influence of noise.

[0075] Specifically, for a random variable X, its information entropy H(X) can be defined as:

[0076]

[0077] Among them, P(xi) is the probability that variable X takes the value xi.

[0078] Therefore, for the environmental data set De, substituting it into the above information entropy calculation formula, the first information entropy obtained is H(De); for the platform state data set Dp, substituting it into the above information entropy calculation formula, the first information entropy obtained is H(Dp).

[0079] According to the environment data set De, the data sequence X can be obtained, and according to the platform status data set, the data sequence Y can be obtained. For the data sequences X and Y, their consistency can be expressed as:

[0080]

[0081] Among them, Cov(X,Y) is the covariance of X and Y, σX and σY are the standard deviations of data series X and data series Y respectively.

[0082] Specifically, assuming that the data sequences corresponding to the sensor data D and Dp are X = {x1, x2, ..., xn} and Y = {y1, y2, ..., yn}, the consistency index ρ(De, Dp) between the two is calculated as follows:

[0083] First calculate the mean:

[0084]

[0085] Next, calculate the covariance:

[0086]

[0087] Calculate the standard deviation:

[0088]

[0089] Finally, calculate the consistency index:

[0090]

[0091] After obtaining the first information entropy, the second information entropy and the consistency index, the probability density P(F|De, Dp) of the pseudo data can be calculated by the improved Bayesian formula to characterize the probability that the data is pseudo data.

[0092] A pseudo data probability density threshold T can be set. If P(F|De, Dp)>T, it is considered that the probability that the data is pseudo data is high.

[0093] A high probability of false data means that there may be more unreliable information in the sensor data. If the probability density is greater than the preset threshold T, it can be determined that the data is affected by noise or sensor failure. Gaussian distribution fusion can take the confidence of the data into account through the weighted average method to reduce the impact of abnormal data on the final fusion result. In this way, it can better maintain stability when facing unreliable data, improve the reliability of data fusion, and ensure that more accurate judgments can still be made in harsh or uncertain environments. Among them, Gaussian distribution fusion refers to the use of the probability density function of the Gaussian distribution to process and combine multi-sensor data.

[0094] Low probability of false data indicates that the sensor data is relatively reliable as a whole, and the consistency between data is high. At this time, all data provided by each sensor can be trusted. Centralized fusion is a fusion method that centralizes all sensor data for processing. If the probability density is less than or equal to the preset threshold T, it can be determined that the impact of false data on the data set is small. Centralized fusion regards all data as equally important and generates the final fusion result through the fusion algorithm. It can make full use of the information of all sensors to generate a comprehensive environment or state perception result, which is conducive to improving data utilization and reducing complex calculation overhead.

[0095] The future motion trend includes the future position, attitude angle, speed and acceleration of the platform. Since the environmental perception feature group includes multiple environmental features and platform state features, the above features reflect the current environmental conditions (such as wind speed, wave height, etc.) and platform state (such as current attitude, speed, acceleration, etc.). Because the motion trend of the platform is directly affected by the environment, the environmental perception feature group provides key information about how the external environment affects the future state of the platform.

[0096] Potential functions are used to connect environmental features with platform state features, indicating the degree of influence of environmental factors on the platform's motion state. Through potential functions, the forces and consequences of environmental changes on the platform can be quantified and simulated. The strength of the potential function is determined based on the degree of influence of environmental features on the platform's state features, which means that by analyzing these potential functions, it is possible to deduce how environmental changes will change the platform's motion trend. By analyzing the current environmental perception feature group, a dynamic model can be established to describe the platform's motion law under current environmental conditions. Dynamic models usually include the platform's dynamic equations and take into account the current inputs (such as environmental forces) and the current state (such as speed and position) to predict the platform's state at future moments.

[0097] The first environmental perception feature group and the second environmental perception feature group of this embodiment come from different data fusion methods (Gaussian distribution fusion or centralized fusion), which makes the final feature group have different focuses. The first feature group emphasizes robustness when uncertainty is high, and the second feature group emphasizes comprehensiveness when data consistency is high. Regardless of which feature group is used, the fused data can more accurately reflect the current environmental status and platform dynamics, thereby providing a more reliable prediction of future motion trends.

[0098] After obtaining the future motion trend of the platform, closed-loop control can be used, that is, the future motion trend is compared with the target state. If there is a deviation, a compensation strategy is adopted to correct the platform in real time, ensuring that the platform can adapt to the dynamically changing environment in real time and remain relatively stable. Specifically, accurate control inputs can be calculated based on future motion trends, such as adjusting the platform's attitude angle or applying a reaction torque. The input is based on an accurate prediction of future trends, thereby ensuring that the platform can respond quickly and effectively to environmental disturbances. Specifically, the force or torque required to maintain the balance of the platform can be calculated based on parameters such as the predicted attitude angle, velocity, and acceleration.

[0099] Since the prediction of future motion trends is based on multi-sensor fusion data, the prediction results are highly reliable because they comprehensively consider information from different sensors. Balancing compensation based on this high-confidence prediction data can effectively improve the accuracy and timeliness of compensation and reduce unexpected motion of the platform caused by environmental disturbances.

[0100] This embodiment can identify and evaluate the probability density of pseudo data by calculating the information entropy and consistency index of environmental data and platform state data. When the probability density of pseudo data is high, the data is processed by Gaussian distribution fusion, thereby reducing the negative impact of pseudo data on the system, improving the accuracy of stability control of the shipborne six-degree-of-freedom stable platform, and reducing the impact on normal activities on the ship. In addition, by adopting two different fusion strategies, Gaussian distribution fusion and centralized fusion, the most suitable data fusion method can be adaptively selected according to the uncertainty and correlation of the data, thereby improving the reliability and accuracy of data fusion, and by processing the fused environmental data and platform state data, the future movement trend of the platform can be more accurately predicted, not only considering the current state of the platform, but also integrating the potential impact of environmental factors on the platform movement, so that effective balance compensation measures can be taken in advance to reduce the occurrence of unstable movement, and the potential function is introduced to connect the environmental characteristics and the platform state characteristics. The strength of the potential function is determined according to the degree of influence of the environmental characteristics on the platform state characteristics, which can model the impact of the environment on the platform state in a more detailed manner, provide a deeper level of feature association, and enhance the response capability of the controller in a complex environment. In summary, by solving the identification and processing of pseudo data, dynamically adjusting the data fusion strategy, accurately predicting future motion trends, and introducing potential function to enhance feature correlation, the uncertainty and instability problems of the traditional shipborne six-degree-of-freedom stabilized platform in complex sea conditions are effectively solved, and the stability and control accuracy of the shipborne six-degree-of-freedom stabilized platform are improved.

[0101] In one implementation of this embodiment, determining the probability density of pseudo data in multi-sensor data according to the first information entropy, the second information entropy and the consistency index includes the following steps:

[0102] S210, calculating the inconsistency index of the environment data and the platform status data according to the consistency index, wherein the sum of the inconsistency index and the consistency index is 1;

[0103] S220, calculating a possibility index of pseudo data according to the first information entropy and the inconsistency index;

[0104] S230 , normalizing the possibility index according to the first information entropy, the second information entropy and the consistency index to obtain the probability density of the pseudo data in the multi-sensor data.

[0105] Based on S130 , the consistency index can be obtained as ρ(De, Dp), and the inconsistency index=1-ρ(De, Dp).

[0106] The possibility index is calculated based on the first information entropy and the inconsistency index. The specific possibility index calculation formula is as follows:

[0107] Possibility index=H(De)·(1-ρ(De,Dp))·σ(De)·|μ(De)-μ(Dp)|·ΔDe·(1-γ(De));

[0108] Where H(De) is the first information entropy; σ(De) is the standard deviation of the environmental data set; |μ(De)-μ(Dp)| is the mean difference between the environmental data set and the platform status data set, which is used to represent the difference between the average values ​​of the two data sets; ΔDe is the rate of change of the environmental data set, which is used to measure the rate of change of the environmental data. A high rate of change indicates that the data changes rapidly and may have more anomalies; γ(De) is the autocorrelation of the environmental data set. 1-γ(De) reflects the correlation of the data over time. If 1-γ(De) is high, it indicates that the data is irregular or unstable and may contain false data.

[0109] The above standard deviation, mean, rate of change and autocorrelation can all be calculated through the data set, and this application will not go into details.

[0110] Normalizing the possibility index means controlling the range of the possibility index within the range of [0, 1]. Specifically, the possibility index is normalized according to the first information entropy, the second information entropy and the consistency index, as shown in the following formula:

[0111]

[0112] In summary, the probability density of pseudo data in multi-sensor data can be obtained.

[0113] The possibility index of this embodiment considers the information entropy of the environmental data because the environmental data De usually reflects external or external factors and may be affected by more external interference and noise, resulting in increased uncertainty and confusion of the data. Therefore, when detecting false data, the information entropy H(De) of the environmental data can more directly reflect the potential abnormalities and false data characteristics in the data.

[0114] This embodiment identifies the deviation between data by calculating the consistency and inconsistency indicators of environmental data and platform status data. Data consistency is a key indicator for determining whether multi-sensor data is reliable. The inconsistency indicator complements the consistency indicator and can effectively distinguish normal data from possible pseudo data. The possibility indicator of pseudo data is calculated based on the first information entropy and the inconsistency indicator. Data with high information entropy usually means greater uncertainty. Combined with the inconsistency indicator, the possibility of a certain data point becoming pseudo data can be evaluated, and then the possibility of pseudo data can be quantified. The pseudo data possibility indicator is normalized using the first information entropy, the second information entropy and the consistency indicator, which is used to convert the pseudo data possibility into a standardized probability density value. In summary, reliable data and pseudo data are effectively distinguished, and uncertainty is better handled during the data fusion process, thereby avoiding the negative impact of pseudo data on the final fusion result, thereby improving the robustness of the entire data fusion process, so that accurate results can still be output when facing unreliable data, ensuring the stability and control accuracy of the platform.

[0115] In one implementation of this embodiment, when the probability density is greater than a preset threshold, Gaussian distribution fusion is performed on the environment data and the platform state data to obtain a first environment perception feature group, including the following steps:

[0116] S310, when the probability density is greater than a preset threshold, extracting features from the environment data to obtain a plurality of environment features E, and extracting features from the platform state data to obtain a plurality of platform state features P;

[0117] S320, using a Gaussian mixture model to fuse the environmental feature E and the platform state feature P to obtain a fused feature group, where the fused feature group includes the environmental feature Ei and the platform state feature Pi;

[0118] S330 , connecting the environmental features Ei and the platform state features Pj in the fused feature group through a potential function to obtain a first environmental perception feature group.

[0119] In order to perform Gaussian distribution fusion on environmental data and platform status data to generate a first environmental perception feature group when the probability density is greater than a preset threshold, this can be achieved through steps such as data preprocessing, Gaussian mixture model input, and potential function connection.

[0120] Specifically, data preprocessing is first performed: 1. Preliminarily remove outliers and noise from the environment data De and platform status data Dp, and standardize the data to have zero mean and unit variance to ensure that data with different features are on the same scale. 2. Extract multiple environmental features E1, E2, ..., E from the environment data and platform data m and multiple platform state features P1, P2, ..., Pn .

[0121] Secondly, when the probability density is greater than a preset threshold, the environment features and platform state features are fused using a Gaussian distribution model to obtain a fused feature group. Fusion can be achieved through weighted averaging, and the weights can be assigned based on the importance or reliability of the features.

[0122] F i =w e E i +w p P i ;

[0123] Among them, Fi is the fused feature, we and wp are the weights of environmental features and platform state features, and they satisfy we+wp=1.

[0124] The fused feature group includes environmental features Ei and platform state features Pj.

[0125] Then, the environmental features Ei and the platform state features Pj in the fused feature group are connected through a potential function to obtain a first environmental perception feature group.

[0126] Specifically, the potential function is used to reflect the degree of influence of environmental characteristics on platform state characteristics. The intensity parameter of the potential function can be estimated through historical data or experimental data to determine the degree of influence of environmental characteristics on platform state characteristics. The degree of influence can be determined using a correlation analysis method. For example, a correlation analysis can be performed by calculating the Pearson correlation coefficient between environmental characteristics and platform state characteristics. The calculated Pearson correlation coefficient can be used as the degree of influence of environmental characteristics on platform state characteristics.

[0127] In this embodiment, the potential function ψ(E i , P j ) is used to connect the environmental features Ei and the platform status features Pj, and reflects the degree of influence of the environmental features on the platform status features, including:

[0128]

[0129] in, and is the mean of the feature, and σ is a tuning parameter that controls the range and strength of the potential function.

[0130] The first environment perception feature group can be constructed by the above-mentioned fusion obtained feature group and potential function. The first environment perception feature group includes multiple environment features E1, E2, ..., E i , multiple platform state features P1, O2, ..., Pj and the corresponding potential function ψ(E i , P j ).

[0131] This implementation can use a Gaussian distribution model to fuse environmental data and platform status data when the possibility of detecting false data is high, and use a potential function to quantify the impact of environmental characteristics on platform status characteristics, effectively improving sensitivity to false data and ensuring that the environmental perception feature group can accurately reflect the real environment and platform status.

[0132] In one implementation of this embodiment, the step of constructing the Gaussian mixture model includes:

[0133] S410, determining K Gaussian components, and initializing the model parameters of the Gaussian mixture model for each Gaussian component;

[0134] S420, using an expectation maximization algorithm to determine final model parameters of the Gaussian mixture model, where the final model parameters include a mixture weight, a mean vector, and a covariance matrix;

[0135] S430, determining the probability density function of each Gaussian component according to the mixing weight, the mean vector and the covariance matrix;

[0136] S440, performing weighted summation on the probability density function of each Gaussian component to determine the overall probability density function of the Gaussian mixture model;

[0137] S450: construct a Gaussian mixture model according to the final model parameters and the overall probability density function.

[0138] In this embodiment, Gaussian distribution fusion adopts a Gaussian mixture model, which can effectively capture the multi-peak distribution characteristics of data and allow fusion of different categories.

[0139] During the construction process, the environmental features and platform status features are first organized into a feature vector in the form of X = {x1, x2, ..., x N}, where each xi is a multidimensional vector containing environmental features and platform state features.

[0140] Then, the number K of Gaussian distributions is determined, that is, K Gaussian components are determined. Specifically, the number K of Gaussian distributions used for modeling is determined, and the model parameters are initialized:

[0141] Mixing weight π k : Initialize the weight of each Gaussian component to satisfy

[0142] Mean vector μ k : Initialized as the mean vector μ of each Gaussian component k ;

[0143] Covariance matrix∑ k: Initialized as the covariance matrix ∑ of each Gaussian component k .

[0144] The probability density function of each Gaussian component is defined as:

[0145]

[0146] Here, d is the dimension of the data.

[0147] The overall probability density function of the Gaussian mixture model is:

[0148]

[0149] Where θ = {π k , μ k ,∑ k} represents all the parameters of the model.

[0150] Afterwards, the parameters of the Gaussian mixture model are estimated by the expectation maximization algorithm. The steps of estimating the final model parameters of the Gaussian mixture model by the expectation maximization algorithm are as follows:

[0151] The expectation-maximization algorithm is an iterative process consisting of two main steps: the expectation step (E step) and the maximization step (M step).

[0152] S1. Expected step: First, calculate the posterior probability (i.e., responsibility) of each sample xi belonging to the Kth Gaussian component:

[0153]

[0154] in, It represents the probability that the data point xi belongs to the Kth Gaussian component at the tth iteration.

[0155] S2, maximization step: Update the model parameters to maximize the log-likelihood function:

[0156] Update blend weights

[0157]

[0158] Update the mean vector

[0159]

[0160] Update the covariance matrix

[0161]

[0162] S3, determine the convergence condition;

[0163] Compute the log-likelihood function:

[0164]

[0165] S4. Determine whether the change of the log-likelihood function logL(θ) is less than a preset threshold (e.g., ∈). If yes, the algorithm converges; otherwise, return to S1 to continue iteration.

[0166] When the algorithm converges, the final parameter θ={π k , μ k ,∑ k} is the best estimated parameter of the Gaussian mixture model.

[0167] Finally, the Gaussian mixture model can be constructed based on the final model parameters and the overall probability density function.

[0168] This embodiment effectively estimates the parameters of the Gaussian mixture model through the expectation maximization algorithm, so that the multi-peak characteristics of the data can be processed, so that the Gaussian mixture model can be used for modeling and analysis of complex data. The expectation maximization algorithm gradually approaches the maximum likelihood estimate by alternately optimizing the expectation and maximization steps, thereby providing an accurate model for the distribution of complex data.

[0169] In one implementation of this embodiment, when the probability density is less than or equal to a preset threshold, the environment data and the platform status data are centrally fused to obtain a second environment perception feature group, including the following steps:

[0170] S510: When the probability density is less than or equal to a preset threshold, extracting features from the environment data to obtain a plurality of environment features E, and extracting features from the platform state data to obtain a plurality of platform state features P;

[0171] S520, dividing the plurality of environmental features E and the plurality of platform status features P into at least two feature groups according to the consistency index, each feature group including the environmental features E and the platform status features P within a preset consistency index range;

[0172] S530, performing feature fusion on the environment feature E and the platform state feature P in each feature group to obtain a fused feature vector f(E, P);

[0173] S540. Combine each feature vector f(E, P) with the corresponding potential function to obtain a second environment perception feature group.

[0174] When the probability density is less than or equal to a preset threshold, the environmental data and the platform status data are centrally fused to obtain a second environmental perception feature group. The specific implementation steps are as follows.

[0175] First, similarly to S310, multiple environmental features E1, E2, ..., E m and multiple platform state features P1, P2, ..., P n .

[0176] Secondly, multiple environmental features E and multiple platform status features P are divided into at least two feature groups. According to the similarity of the features (the consistency index between the environmental features and the platform status features is within the preset consistency index range), the features are divided into several groups, each group containing similar environmental features and platform status features.

[0177] Then, the environment features and platform status features are centrally fused, that is, the environment features and platform status features are combined through the fusion strategy to form a new feature vector F: F = f(F e , F p ).

[0178] Among them, J(·) is the feature fusion function, which can be nonlinearly combined through machine learning methods.

[0179] S330 has introduced the construction of the potential function in detail, and this application will not repeat it again.

[0180] Afterwards, each group of fused features F is compared with the corresponding potential function ψ(E i , P j ) to form the second environmental perception feature group:

[0181] The second environmental perception feature group = {F 1 , F 2 , ..., φ 1 ,φ 2 , ...};

[0182] Among them, F i represents the fusion features of the i-th group of data, φ i represents the corresponding potential function.

[0183] This implementation can effectively process data in low probability density areas and generate a second environmental perception feature group through centralized fusion. The fused feature vector is combined with the potential function to obtain the second environmental perception feature group. By combining the potential function (which represents the intensity of the impact of the environment on the platform state) with the fused feature vector, the impact of the environment on the platform state can be better understood and characterized, thereby generating a more accurate environmental perception feature group. The introduction of the potential function provides a quantitative expression for the relationship between environmental features and platform state features, further enhancing the perception of complex environments.

[0184] In one implementation of this embodiment, feature fusion is performed on the environment feature E and the platform state feature P in each feature group to obtain a fused feature vector f(E, P), including the following steps:

[0185] S610, normalizing the environment feature E and the platform state feature P so that the value ranges of the environment feature E and the platform state feature P are consistent;

[0186] S620. Input the normalized environmental features E and platform state features P into a pre-trained multi-layer perceptron neural network to obtain a fused feature vector f(E, P), wherein the multi-layer perceptron neural network includes an input layer, a hidden layer, and an output layer.

[0187] First, the environment feature E and the platform state feature P are preprocessed to make them suitable for the nonlinear model. Specifically, the environment feature E and the platform state feature P are normalized to make the value range of the environment feature E and the platform state feature P consistent.

[0188] Afterwards, feature fusion is performed through a nonlinear model. In this embodiment, the nonlinear model is a multilayer perceptron neural network. The normalized environmental features E and platform state features P are input into the pre-trained multilayer perceptron neural network to obtain the fused feature vector f(E, P).

[0189] Specifically, the input layer is used to take the preprocessed environmental features and platform state features as input and connect them together to form a vector. Several hidden layers are used, and each layer uses a nonlinear activation function (such as ReLU) to perform a nonlinear transformation on the input to capture the complex nonlinear relationship between environmental features and platform state features. The output layer is used to output the fused feature vector, which contains the nonlinear combination information of environmental features and platform state features.

[0190] This embodiment can more accurately characterize the impact of the environment on the platform through normalization processing and multi-layer perceptron neural network, and can more accurately perform environmental perception and state prediction based on fused feature vectors.

[0191] In one implementation of this embodiment, determining the future motion trend of the shipborne six-degree-of-freedom stabilized platform by using the first environment perception feature group or the second environment perception feature group includes the following steps:

[0192] S710. Input the first environment perception feature group or the second environment perception feature group into a preset six-degree-of-freedom rigid body dynamics equation, and use a numerical integration method to solve the six-degree-of-freedom rigid body dynamics equation to obtain the future position, attitude angle, velocity and acceleration of the shipborne six-degree-of-freedom stable platform. The future position, attitude angle, velocity and acceleration are used to describe the future movement trend.

[0193] The six-degree-of-freedom rigid body dynamics equation is used to describe the linear and angular motion of the shipborne platform. The equation consists of two parts:

[0194] 1. Linear momentum balance equation:

[0195] MV+C(V)V+D(V)V+g(η)=τ;

[0196] Where M is the inertia matrix (including the attached mass matrix); C(V) is the Coriolis and centrifugal force matrix; D(V) is the damping matrix; g(η) is the gravity and buoyancy matrix; τ is the external force and external moment.

[0197] 2. Angular momentum balance equation:

[0198]

[0199] Where η is the position and attitude angle vector (including three-dimensional position x, y, z and Euler angles φ, θ, ψ); J(η) is the transformation matrix related to the attitude angle; V is the linear velocity and angular velocity vector.

[0200] The first environment perception feature group or the second environment perception feature group is mapped to the external force and external moment vector τ. Specifically, the environment feature can be converted into the corresponding force and moment through a preset external force and moment model.

[0201] In order to solve the six-degree-of-freedom rigid body dynamics equation, a numerical integration method may be selected. The numerical integration method may be a fourth-order Runge-Kutta method.

[0202] The numerical integration solution step includes: Step 1. Setting the initial position η0 and the initial velocity V0. Step 2. Using the first environment perception feature group or the second environment perception feature group to calculate the external force and torque τ(t) at the current moment. Step 3. Calculating the instantaneous linear acceleration V(t) and angular acceleration according to the six-degree-of-freedom rigid body dynamics equation. Step 4. Update the position and velocity using the Runge-Kutta method.

[0203] Specifically, the linear acceleration:

[0204]

[0205] Angular acceleration:

[0206] Update position and velocity using Euler's method, including:

[0207] For any variable y(t):

[0208] k1=f(t n ,y n );

[0209]

[0210] k4=f(t n +h,y n +hk3);

[0211]

[0212] Where h is the time step and f(t,y) represents the derivative.

[0213] For linear velocity and angular velocity V:

[0214]

[0215] For position and attitude angle η:

[0216]

[0217] Repeat steps 2 to 4 and advance the time step h until the required prediction time range is reached, and the position, attitude angle, linear velocity and linear acceleration at each time step can be obtained as the future motion trend of the carrier platform.

[0218] This embodiment can accurately predict the future motion trend of the shipborne six-degree-of-freedom stable platform by mapping the first or second environmental perception feature group into an external torque input into the six-degree-of-freedom rigid body dynamics equation and solving it using the numerical integration method, thereby ensuring the accuracy of the future motion trend.

[0219] In one implementation of this embodiment, the balance compensation of the shipborne six-degree-of-freedom stable platform is performed according to the future movement trend, including the following steps:

[0220] S810, calculating a position error according to the future position and the acquired target position;

[0221] S820, calculating a speed error according to the speed and the acquired target speed;

[0222] S830, calculating an acceleration error according to the acceleration and the acquired target acceleration;

[0223] S840, calculating an attitude angle error according to the attitude angle and the acquired target attitude angle;

[0224] S850, generating a control signal according to the position error, the velocity error, the acceleration error and the attitude angle error, so as to compensate the balance of the shipborne six-degree-of-freedom stabilization platform through the control signal.

[0225] In order to compensate for the balance according to the future movement trend of the shipborne six-degree-of-freedom stable platform and ensure that it remains stable during movement, it is usually necessary to control the position, speed and acceleration of the driving axis of the platform. The specific implementation methods are as follows:

[0226] First, by predicting the future motion trend of the platform, the key motion parameters that need to be compensated are determined, including position, attitude angle, linear velocity and linear acceleration. By comparing the current position and the target position, the position error is calculated. After the position error is obtained, the precise position control of the drive axis is performed through the PID controller.

[0227] In one embodiment, the control law of the PID controller can be:

[0228]

[0229] Where: K p is the proportional gain; K i is the integral gain; K d is the differential gain; u η (t) is the control output, which is used to adjust the drive shaft.

[0230] The output u of the PID controller η (t) is directly used to adjust the position of the drive shaft to ensure the target position accuracy of the platform.

[0231] Similarly, the speed error is calculated based on the target speed and the current speed. The calculated speed error can be used to adjust the control signal of the drive shaft speed through the PID controller. The output of the PID controller will be directly used to adjust the speed of the drive shaft to ensure that the platform can move at the target speed.

[0232] Similarly, the acceleration error is calculated and can be compensated by acceleration feedback control, that is, by adjusting the torque or thrust of the drive shaft.

[0233] Attitude angle control aims to keep the platform's attitude angle (roll angle, pitch angle, yaw angle) within a predetermined range, or to dynamically adjust it according to the predicted future attitude angle target.

[0234] Similar to the control of position, velocity and acceleration, it is necessary to first calculate the error between the current attitude angle and the target attitude angle. In this embodiment, the PID controller for each attitude angle is different.

[0235] Finally, the attitude angle control is combined with the linear motion control. The final control signal can be a linear combination of the outputs of different controllers to adjust the effects of linear and rotational motion.

[0236] According to the final control signal, the control signal can be applied to the drive shaft to adjust the position, speed, acceleration and attitude angle, so as to compensate the balance of the shipborne six-degree-of-freedom stable platform through the control signal.

[0237] This implementation method can better cope with external disturbances and ensure the stability of the platform in complex sea conditions by incorporating attitude angle control into the balance compensation strategy of the six-degree-of-freedom platform. By integrating the control of position, speed, acceleration and attitude angle and using the state feedback mechanism, precise control of the shipborne six-degree-of-freedom stable platform can be achieved.

[0238] The present application also provides a controller, including:

[0239] a memory configured to store instructions; and

[0240] The processor is configured to call instructions from the memory and implement the above-mentioned multi-sensor fusion method based on the shipborne six-degree-of-freedom stable platform when executing the instructions.

[0241] Figure 2 FIG. 1 shows a schematic diagram of the structure of a multi-sensor fusion system based on a shipborne six-degree-of-freedom stable platform provided in an embodiment of the present application. Figure 2 As shown, the embodiment of the present application also provides a multi-sensor fusion system based on a shipborne six-degree-of-freedom stable platform, including:

[0242] Controller 100;

[0243] The shipborne six-degree-of-freedom stable platform 200 is connected to the controller 100 .

[0244] An embodiment of the present application also provides a machine-readable storage medium, on which instructions are stored, and the instructions are used to enable a machine to execute the above-mentioned multi-sensor fusion method based on a shipborne six-degree-of-freedom stable platform.

[0245] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code.

[0246] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0247] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0248] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0249] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0250] The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. The memory is an example of a computer-readable medium.

[0251] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0252] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0253] The above are only embodiments of the present application and are not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included within the scope of the claims of the present application.

Claims

1. A multi-sensor fusion method based on a shipborne six-degree-of-freedom stable platform, characterized in that: include: Acquiring multi-sensor data through a multi-sensor disposed on a shipborne six-degree-of-freedom stable platform, wherein the multi-sensor data includes environmental data and platform status data; respectively calculating a first information entropy of the environment data and a second information entropy of the platform state data to quantify the uncertainty of the environment data and the platform state data; Calculating a consistency index between each of the environment data and each of the platform status data to quantify the correlation between the environment data and the platform status data; Determining a probability density of pseudo data in the multi-sensor data according to the first information entropy, the second information entropy and the consistency index; When the probability density is greater than a preset threshold, performing Gaussian distribution fusion on the environment data and the platform state data to obtain a first environment perception feature group; When the probability density is less than or equal to a preset threshold, the environment data and the platform status data are centrally integrated to obtain a second environment perception feature group; determining a future motion trend of the shipborne six-degree-of-freedom stabilized platform by using the first environment perception feature group or the second environment perception feature group; Performing balance compensation on the shipborne six-degree-of-freedom stable platform according to the future motion trend; The first environment perception feature group or the second environment perception feature group includes a plurality of environment features, a plurality of platform state features and a plurality of potential functions, wherein the potential function is used to connect the environment features with the platform state features, and the strength of the potential function is determined based on the influence of the environment features on the platform state features; When the probability density is greater than a preset threshold, Gaussian distribution fusion is performed on the environment data and the platform state data to obtain a first environment perception feature group, including: When the probability density is greater than a preset threshold, feature extraction is performed on the environment data to obtain a plurality of the environment features E, and feature extraction is performed on the platform state data to obtain a plurality of the platform state features P; The environment feature E and the platform state feature P are fused using a Gaussian mixture model to obtain a fused feature group, wherein the fused feature group includes the environment feature Ei and the platform state feature Pj; Connecting the environment feature Ei and the platform state feature Pj in the fused feature group through a potential function to obtain the first environment perception feature group; When the probability density is less than or equal to a preset threshold, the environment data and the platform status data are centrally integrated to obtain a second environment perception feature group, including: When the probability density is less than or equal to a preset threshold, feature extraction is performed on the environment data to obtain a plurality of the environment features E, and feature extraction is performed on the platform state data to obtain a plurality of the platform state features P; According to the consistency index, the plurality of environmental features E and the plurality of platform state features P are divided into at least two feature groups, each of the feature groups including the environmental features E and the platform state features P within a preset consistency index range; Performing feature fusion on the environment feature E and the platform state feature P in each feature group to obtain a fused feature vector f(E, P); Combining each of the feature vectors f(E, P) with a corresponding potential function to obtain a second environment perception feature group; Determining the future movement trend of the shipborne six-degree-of-freedom stabilized platform by using the first environment perception feature group or the second environment perception feature group includes: Inputting the first environment perception feature group or the second environment perception feature group into a preset six-degree-of-freedom rigid body dynamics equation, and solving the six-degree-of-freedom rigid body dynamics equation by a numerical integration method to obtain a future position, attitude angle, velocity and acceleration of the shipborne six-degree-of-freedom stable platform, wherein the future position, attitude angle, velocity and acceleration are used to describe the future motion trend; The performing balance compensation on the shipborne six-degree-of-freedom stable platform according to the future motion trend includes: Calculating a position error based on the future position and the acquired target position; Calculating a speed error based on the speed and the acquired target speed; Calculating an acceleration error according to the acceleration and the acquired target acceleration; Calculating an attitude angle error according to the attitude angle and the acquired target attitude angle; A control signal is generated according to the position error, the velocity error, the acceleration error and the attitude angle error, so as to perform balance compensation on the shipborne six-degree-of-freedom stable platform through the control signal.

2. The method according to claim 1, characterized in that The determining, according to the first information entropy, the second information entropy and the consistency index, the probability density of the pseudo data in the multi-sensor data includes: Calculating, according to the consistency index, an inconsistency index of the environment data and the platform status data, wherein the sum of the inconsistency index and the consistency index is 1; Calculating a possibility index of pseudo data according to the first information entropy and the inconsistency index; The possibility index is normalized according to the first information entropy, the second information entropy and the consistency index to obtain a probability density of pseudo data in the multi-sensor data.

3. The method according to claim 1, characterized in that The steps of constructing the Gaussian mixture model include: Determine K Gaussian components, and initialize the model parameters of the Gaussian mixture model for each of the Gaussian components; The final model parameters of the Gaussian mixture model are determined by using an expectation maximization algorithm, wherein the final model parameters include a mixture weight, a mean vector, and a covariance matrix; Determine a probability density function of each of the Gaussian components according to the mixing weight, the mean vector and the covariance matrix; Performing weighted summation on the probability density function of each Gaussian component to determine the overall probability density function of the Gaussian mixture model; A Gaussian mixture model is constructed according to the final model parameters and the overall probability density function.

4. The method according to claim 1, characterized in that: The step of fusing the environment feature E and the platform state feature P in each feature group to obtain a fused feature vector f(E, P) includes: Normalizing the environment feature E and the platform state feature P so that the value ranges of the environment feature E and the platform state feature P are consistent; The normalized environmental feature E and the platform state feature P are input into a pre-trained multi-layer perceptron neural network to obtain a fused feature vector f(E, P), wherein the multi-layer perceptron neural network includes an input layer, a hidden layer and an output layer.

5. A controller, characterized in that: include: a memory configured to store instructions; as well as A processor is configured to call the instructions from the memory and implement the multi-sensor fusion method based on a shipborne six-degree-of-freedom stable platform according to any one of claims 1 to 4 when executing the instructions.

6. A multi-sensor fusion system based on a shipborne six-degree-of-freedom stabilized platform, characterized in that: include: A controller according to claim 5; A shipborne six-degree-of-freedom stabilized platform is connected to the controller.

Images