Unmanned ship positioning method based on multi-sensor data fusion

Abstract

An unmanned ship positioning method based on multi-sensor data fusion, the steps of which are as follows: first, pre-process data collected by an unmanned ship multi-sensor positioning system, then perform a reliability judgment on the unmanned ship multi-sensor positioning data through a reliability distance test, and assign a corresponding reliability factor to the tested positioning data, perform a test compensation for faulty data using a consistency test and variance weighting, and perform a filtering process on each sensor positioning data after testing and weighting based on basic particle filtering to realize data augmentation, and finally perform a fusion filtering output process on the unmanned ship multi-sensor positioning data using a new threshold hierarchical particle filtering algorithm to obtain accurate unmanned ship navigation trajectory positioning information. The method improves the fault tolerance performance and algorithm relevance of the unmanned ship multi-sensor positioning system, ensures the reliability of the sensor positioning data, and achieves the purpose of accurately positioning the unmanned ship navigation trajectory.

Description

Technical Field

[0001] The present invention belongs to the technical field of positioning navigation, relates to a fusion processing technology based on multi-sensor information collection, and particularly relates to an unmanned ship positioning method based on multi-sensor data fusion.

Background Art

[0002] The development and application of an intelligent unmanned ship positioning system supported by modern sensor information technology can not only promote the development of the unmanned surface ship in the fields of offshore patrol, water quality monitoring, aquaculture, etc., but also greatly reduce the work intensity of workers and improve work efficiency.

[0003] In order to realize the efficient and accurate operation of the unmanned ship, it first depends on whether the obtained sensor positioning information can accurately sense the position of the unmanned surface ship, that is, whether the unmanned ship can use the obtained positioning information to ensure its work efficiency and accuracy. With the development of the automatic navigation technology of the unmanned surface ship, the requirements for the positioning accuracy and stability of the unmanned ship are increasing. Due to the influence of the complex working environment on the water surface and the shore, the sensor information may be lost and the positioning may become inaccurate. Therefore, there are great disadvantages when using a single sensor, and in the case of multi-sensor data fusion, their advantages are utilized complementarily to effectively overcome the above disadvantages, which has become the development trend of the combined positioning system. Therefore, research and application of the multi-sensor data fusion algorithm of the unmanned ship positioning system are carried out, which helps to improve the fault tolerance performance of the algorithm, and further realizes the high-efficiency filtering process of the unmanned ship position data by the multi-sensor data fusion algorithm, and finally achieves the accurate positioning of the unmanned ship position and the improvement of the production operation efficiency.

Summary of the Invention

Problems to be Solved by the Invention

[0004] The object of the present invention is to provide a positioning method for an unmanned ship based on multi-sensor data fusion for the multi-sensor data fusion positioning of the unmanned ship navigation trajectory in order to overcome the major drawbacks of the prior art by using a single sensor.

Means for Solving the Problems

[0005] In order to achieve the above object, the present invention is realized by using the following technical solutions.

[0006] A positioning method for an unmanned ship based on multi-sensor data fusion, which first preprocesses the positioning data collected by the unmanned ship multi-sensor positioning system, and then performs a reliable distance determination on the unmanned ship positioning data and assigns a corresponding reliability factor. At the same time, a fault inspection and weighted compensation are performed on the positioning data, and a filtering augmentation is performed on the positioning data. Finally, a multi-sensor data fusion filtering output is performed by using a data fusion algorithm, thereby realizing the positioning of the unmanned ship navigation trajectory. The specific steps are as follows: Step 1. Data preprocessing, Preprocess the unmanned ship positioning data collected by the unmanned ship multi-sensor positioning system.

[0007] Step 2. Data reliability determination and assignment, The data reliability determination and assignment include performing a reliability determination and assigning a reliability factor based on the data collected by the multi-sensor.

[0008] Step 3. Data fault inspection and compensation, The data fault inspection and compensation include performing a consistency inspection on the multi-sensor data and performing a weighted compensation process on the inconsistent data collected by the positioning system.

[0009] Step 4. Data augmentation, The data augmentation includes performing filtering processes on the pre-processed and inspected multi-sensor data and the compensated multi-sensor data respectively based on a fundamental particle filtering algorithm.

[0010] Step 5. Data fusion The data fusion includes constructing a Gaussian mixture model, setting a hierarchical sampling ratio capacity related to an adaptive threshold and a confidence factor, designing a new threshold hierarchical particle filtering algorithm, and further performing fusion filtering on the unmanned ship navigation trajectory positioning data to output unmanned ship navigation trajectory positioning information.

[0011] Furthermore, the specific content and method steps of the data preprocessing described in Step 1 for preprocessing the position data collected by the unmanned ship multi-sensor positioning system are as follows: A) Based on the longitude and latitude information of the unmanned ship navigation trajectory collected by the positioning system sensor, unify the time and space reference for it.

[0012] B) By constructing a communication environment that satisfies the effective connection of multiple nodes of the positioning system according to the river channel where the unmanned ship sails, ensure that the blind node located on the unmanned ship is within a network consisting of four signal receiving nodes with known transmission signal strength and position coordinates, and thereby calculate and collect the position coordinates when the unmanned ship sails through the river channel.

[0013] C) Based on the Gauss-Krüger projection principle, transform and unify the multi-sensor positioning system coordinates.

[0014] Furthermore, the data reliability determination and assignment described in Step 2 include performing a reliability determination and performing a confidence factor assignment based on the data collected by the multi-sensor. The specific content and method steps of the data reliability determination are as follows: 1) According to the measurement model p i (x) of the measurement data of the unmanned ship positioning system, calculate the reliability distance d i between the sensor data, and the p i(x), d i The calculation formula of is as follows,

Number

Number

Number

[0015] 2) According to the Gaussian probability model, the d obtained in step 1) i is rewritten as the metric p r (Z i ) in the probabilistic sense between sensor measurement data, the sensor support confidence level is set, and thereby the reliability of different sensor information is judged. The calculation formula of the p r (Z i ) is as follows,

Number

[0016] 3) According to the magnitude of the information metric value in the probabilistic sense between measurement data, the determined positioning data is divided into different confidence intervals so that the reliable data approaches the region with a high confidence level. Also, the set confidence level ε is the sensor measurement variance average value

Number

Number

[0017] Furthermore, the specific content and method steps for allocating the data confidence factor in the data confidence determination and allocation described in Step 2 are as follows: (A) Construct the norm equation of the dynamic support factor β i (k) of the measurement information collected by the multi-sensor at time k and the Gaussian probability measurement model p i (k), and the calculation formula of the β i (k) is as follows:

Number

[0018] (B) Based on the multi-sensor dynamic support factor β i (k) calculated in Step (A), calculate the measurement error w i of the system, and the calculation formula of the w i is as follows: w i = z i - Aβ i (k) In the formula, A is the state transition matrix, and z i is the measurement value of the unmanned ship multi-sensor positioning system, where i = 1, 2, …, n.

[0019] (C) Calculate the variance of the measurement error w i of the system as

Number

Number

[0020] According to the calculated reliability of the measurement data of the unmanned ship multi-sensor, the corresponding reliability factor is assigned, and the reliability factor is associated with the new threshold hierarchical particle filtering algorithm. The higher the reliability of the measurement data, the higher the corresponding fusion bias, which improves the processing accuracy of the unmanned ship multi-sensor data fusion algorithm.

[0021] Furthermore, the data fault inspection and compensation described in step 3 includes performing a consistency check on the multi-sensor data and performing a weighted compensation process on the inconsistent fault data collected by the positioning system. The specific content and method of the consistency check adopt the following steps: (1) Arithmetically average the sensor measurement data obtained by the unmanned ship positioning system to obtain the arithmetic mean value

Number

[0022] Furthermore, in the data failure inspection compensation described in step 3, the method of performing weighted compensation processing on the inconsistent failure data specifically adopts the following steps: (I) The arithmetic mean value of the multi-sensor measurement data of the unmanned ship [Number] As the offset-free estimated value of [], when the unmanned ship multi-sensor positioning system measures the unmanned ship navigation trajectory at different positions in the same space, the information measurement variance of the i-th sensor system [Number] (II) According to the data set recorded by the unmanned ship multi-sensor positioning system after performing m measurements on the unmanned ship, the j-th measurement data of the i-th sensor is x ij Let it be, and replace the sensor measurement information x ij in step (I) with x i to obtain the data set information variance obtained by multiple measurements [Number] In the formula, i = 1, 2,..., n and j = 1, 2,..., m (III) Based on the position information data set variance collected by the unmanned ship multi-sensor positioning system calculated in step (II), calculate the fusion weight k i of the inconsistent failure data. The calculation formula of the k i is as follows [Number] (IV) Based on the obtained fusion weight k i , perform variance weighted compensation on the inconsistent failure data [Number] to obtain sensor measurement data that meets the requirements of the basic particle filtering process

Number

[0023] Furthermore, the data augmentation described in step 4 includes performing filtering processes on the pre - processed and inspected, compensated multi - sensor data respectively based on the particle filtering algorithm. The specific content and method steps of the said data augmentation are as follows: (i) Establish the particle filtering model of the unmanned ship multi - sensor positioning system, substitute the unmanned ship position information after consistency inspection and variance weighting into the system model, and describe the state and measurement models of the sensor system by the following equations

Number

Number

Number

Number

[0024] Particle propagation samples the unmanned ship multi-sensor positioning system state transition model [Number] After sampling, a new particle state \(X\) k is generated. Here, \(x\) k-1 is the particle state after the previous resampling, and \(Z\) k is the detection data of the sensor system. The measurement data of the unmanned ship multi-sensor positioning system is filtered using the particle filtering algorithm to realize data augmentation, reduce the interference of measurement data by environmental noise, and improve the fusion accuracy of the unmanned ship multi-sensor measurement data.

[0025] Furthermore, the data fusion described in step 5 includes constructing a Gaussian mixture model, setting the hierarchical sampling ratio capacity related to the adaptive threshold and the confidence factor, designing a new threshold hierarchical particle filtering algorithm, and further performing fusion filtering on the unmanned ship navigation trajectory positioning data to output the unmanned ship navigation trajectory positioning information. The specific content and method steps of constructing the Gaussian mixture model are as follows: (a) Extract the multi-sensor data set processed by the particle filtering algorithm, and extract the Sigma point set [Number] is the set of Sigma points after unscented transformation, where n a is the dimension of the Sigma points, λ is the scale parameter, and (b) Fuse the latest measurement information with the obtained Sigma sampling point set, and the system state [Number] In the formula, k k is the Kalman gain, z k is the measurement information, and [Number] is the covariance measured by the weighted Sigma point set, and (c) Construct the unmanned ship multi-sensor positioning system state obtained in step (b) [Number] and sample from the constructed proposal distribution, (d) According to the Gaussian mixture model, the posterior probability density function with a time step size of k [Number] is the i-th component in the Gaussian mixture model, C(k) is the number of component units of the discrete samples, and ξ is the discrete point component weight, and (e) Fuse the discrete sampling points sampled in step (c) and their corresponding weights [Number] into the Gaussian mixture component units in step (d), and the constructed continuous probability density function [Number] In the formula, p(k) is the covariance of the discrete particle filtering distribution,

Number

Number

[0026] First, represent the posterior probability density function representing the state with a continuous probability density function constructed by discrete particles, then approximate the constructed continuous probability density function with a Gaussian mixture distribution, and extract particle samples instead of resampling on it to maintain particle diversity, avoid sample deficiency, and improve filtering fusion accuracy.

[0027] Furthermore, the specific content and method steps of setting the adaptation threshold in the data fusion described in step 5 are as follows: a) After selecting the importance sampling process, the particle x with the maximum weight in the discrete particle sample set c is used as the clustering center, and the martensite distance D i between it and other particles i is calculated. The calculation formula of the D i is as follows:

Number

Number

Equation

Equation

Equation

[0028] The higher the probability density of the particle, the smaller the D i calculated in step a). The adaptive threshold value T cSince it is directly proportional to the particle probability density covariance, D in step e) i compared with T c is D i can adjust its size according to the size of, when clustering Gaussian mixtures, there is no need to continuously adjust the threshold to integrate similar units, reducing the number of clustering times for other particle samples and improving the clustering efficiency. In addition, since the constructed adaptive threshold is related to the particle probability mass, when integrating similar units for a particle set, based on the size of the threshold, particles with similar weight sizes are clustered into one component unit, ensuring that the difference value of the weights between the particle sets within the same component unit is minimized. After integrating similar units using the weighted point set mixed Gaussian distribution constructed in step (e), it is approximated by the posterior probability density function to improve the sampling accuracy in the resampling process.

[0029] Furthermore, in the data fusion described in step 5, the specific content and method steps of associating the reliability factor with the hierarchical sampling ratio capacity and further performing fusion filtering on the unmanned ship navigation trajectory positioning data to output the unmanned ship navigation trajectory positioning information are as follows: (1) Based on the hierarchical theory, each probability density function is a continuous probability density function p(x)

Number

[0030] (2) Let the ratio capacities of the number of particles in each of the l a 、l b 、l c layers be N / 4, N / 3, N / 3.

[0031] (3) Substitute the reliability factor into the weight optimization combination for calculation.​

[0032] (4) l b 、l c The weights in the layer are the average value [Number] Perform a weight optimization combination for particles smaller than, and the weights of the optimized particles [Number] are obtained, and hierarchical sampling is performed on the sample data, and the [Number] (5) Obtain the multi-sensor data fusion sampling result obtained in step (4).

[0033] (6) Output the data fusion sampling result obtained in step (5) from the blind node mounted on the unmanned ship in the form of a log file.

[0034] (7) By networking the PC-side coordinator node arranged on the riverbed with the unmanned ship blind node, the unmanned ship position information output by the blind node in step (7) is acquired in real time, thereby realizing the positioning of the unmanned ship navigation trajectory.

[0035] Hierarchicalize the sampling samples of the continuous probability density function of the constructed weighted point set, and set the ratio capacity of each sampling layer to ensure that the allocation of the sampling particle number in the layer is reasonable, and then l a sample the layer set, and then l b 、l cOptimize the combination for the weights of the particles in the layer group, increase its probability mass, and associate the measurement data reliability factor collected by the unmanned ship multi-sensor positioning system with the hierarchical sampling in the multi-sensor data algorithm. When performing data fusion on the unmanned ship multi-sensor measurement data using the new threshold hierarchical particle filtering algorithm, preferentially sample and fuse the sensor measurement values with large reliability factors.

Advantages of the Invention

[0036] The present invention has the following advantages and beneficial effects.

[0037] (1) In the present invention, information supplementation is performed using the positioning data collected by the unmanned ship multi-sensor, effectively overcoming the problem of sensor signal loss due to environmental signal interruption, and ensuring the effectiveness of the data collected by the sensor.

[0038] (2) In the present invention, consistency checking is performed on the unmanned ship position information collected by the multi-sensor positioning system, and weighted compensation is performed on the fault data, thereby improving the fault tolerance performance of the data fusion algorithm and ensuring the reliability of the particle set samples sampled by the basic particle filtering algorithm.

[0039] (3) Solve the integral operation in Bayesian estimation according to the Monte Carlo method adopted by the basic particle filtering, remove environmental noise interference, ensure the processing accuracy of the sampling sample data set, and further improve the reliability of the unmanned ship multi-sensor measurement data.

[0040] (4) In the present invention, in the importance sampling process of the data fusion algorithm, the current measurement information is fused into the particle set proposal distribution, the proposal distribution is made closer to the actual posterior probability density, the estimation performance of the algorithm is improved. Also, an adaptive threshold is constructed in the clustering analysis of the Gaussian mixture unit, discrete particle samples are integrated into similar component units, the complexity of the clustering operation is reduced, and the real-time performance and operation efficiency of the signal processing of the system are improved.

[0041] (5) In the present invention, the sampling samples of the continuous probability density function of the constructed weighted point set are hierarchized, and the ratio capacity of each sampling layer is set to ensure that the allocation of the sampling particle number in the hierarchy is reasonable. Then, the l a layer set is sampled, and for the weights of the particles in the l b and l c layer sets, an optimization combination is performed to increase its probability mass. Also, the measurement data reliability factor collected by the unmanned ship multi-sensor positioning system is associated with the hierarchical sampling in the multi-sensor data algorithm. When performing data fusion on the unmanned ship multi-sensor measurement data using the new threshold hierarchical particle filtering algorithm, the sensor measurement values with large fusion reliability factors are preferentially sampled, further increasing the reference value of the entire information sample, and finally improving the fusion positioning accuracy of the unmanned ship position data by the data fusion algorithm.

Brief Description of the Drawings

[0042]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7

Figure 8

Figure 9

Figure 10(a)

Figure 10(b)

Figure 11

Figure 12

Embodiments for Carrying out the Invention

[0043] To make the objectives and technical solutions of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below in connection with the drawings of the embodiments of the present invention. Obviously, the described embodiments are only some, rather than all, of the embodiments of the present invention. Based on the described embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative labor shall fall within the protection scope of the present invention.

[0044] As shown in Figure 1, a method for positioning an unmanned ship based on multi-sensor data fusion of the present invention, first preprocessing the positioning data collected by the unmanned ship multi-sensor positioning system, and performing a reliability distance determination on the unmanned ship positioning data and assigning corresponding reliability factors, at the same time performing a fault inspection and weighted compensation on the positioning data, and performing a filtering process on the positioning data to achieve data augmentation, and finally using a new threshold hierarchical particle filtering algorithm to perform multi-sensor data fusion filtering output, thereby realizing accurate positioning of the unmanned ship navigation trajectory.

[0045] Specifically, the following steps are adopted: (1) Data preprocessing (2) Data reliability determination and assignment (3) Data fault inspection and compensation (4) Data augmentation (5) Data fusion

[0046] As shown in Figure 2, the method for preprocessing the positioning data of the unmanned ship collected by the unmanned ship multi-sensor positioning system includes the following steps.

[0047] (A) Based on the longitude and latitude information of the unmanned ship's navigation trajectory collected by the positioning system sensor, perform the unification of the time and space reference for it.

[0048] (B) According to the river channel where the unmanned ship sails, construct a communication environment that satisfies the effective connection of the multi-nodes of the positioning system, so that the blind node located on the unmanned ship is within a network composed of 4 signal receiving nodes with known transmission signal strength and position coordinates, thereby calculating and collecting the position coordinates when the unmanned ship sails through the river channel.

[0049] (C) Based on the Gauss-Kruger projection principle, convert and unify the multi-sensor positioning system coordinates.

[0050] Use the positioning data collected by the unmanned ship multi-sensor to perform information supplementation, effectively overcome the problem of sensor signal loss caused by environmental signal interruption, and ensure the effectiveness of the data collected by the sensor.

[0051] As shown in Figure 3, the method for performing reliability determination and assigning reliability factors based on the data collected by the multi-sensor includes the following steps: 1) According to the measurement data collected by the positioning system, construct the multi-sensor measurement model p i (x), and calculate the reliability distance d i between sensor data, and the calculation of the p i (x), d i is as follows: [Number] In the formula, x i is the measured value of the i-th sensor, μ is the true value of the measurement feature, and θ i is the measurement accuracy of the i-th sensor information, and σ i is the measurement error of the i-th sensor information. [Number]

[0052] In the formula, [Number] Each of them is the measured value collected by the multi-sensor at the i-th moment, [Number] is the average value of the measurement variance, Z is a random variable following the standard normal distribution, and i = 1, 2,..., n.

[0053] 2) According to the Gaussian probability model, the d obtained in step 1) i is rewritten as the metric p r (Z i ) in terms of probability between sensor measurement data, the sensor support confidence level is set, thereby judging the reliability of different sensor information, and the calculation of the p r (Z i ) is as follows: [Number] In the formula, Z i is the measurement data of the multi-sensor, ε is the confidence level, and K is the variable coefficient of the sampling sample probability interval.

[0054] 3) Judge the magnitude of the reliability of the position information collected by the multi-sensor according to the sensor reliability distance, and divide the positioning data into different confidence intervals based on the magnitude of the information reliability.

[0055] 4) Dynamic support factor β of measurement information collected by multiple sensors at time k i (k) and the Gaussian probability measurement model p i (k) norm equation is constructed, and the β i The calculation of (k) is as follows:

number

[0056] 5) The multi-sensor dynamic support factor β obtained in step (4) i Based on (k), the measurement error of the system w i Calculate the w i The calculation of is as follows: w i =z i -Aβ i (k) In the formula, A is the state transition matrix, and z i are the measurements of the unmanned ship multi-sensor positioning system, for i = 1, 2, ..., n.

[0057] 6) System measurement error w i The variance of

number

number

[0058] According to the magnitude of the probability semantic information metric value between the measurement data, the determined positioning data is divided into different confidence intervals, so that the reliable data approaches the high confidence region, and the set confidence level ε is determined by the sensor measurement variance average value

number

[0059] As shown in FIG. 4, the method of performing weighted compensation processing on the inconsistent fault data collected by the multi-sensor by performing a consistency check on the multi-sensor data includes the following steps: (I) Arithmetically average the sensor measurement data obtained by the unmanned ship positioning system to obtain the arithmetic mean value [Number] In the formula, x i is the sensor measurement information, where i = 1, 2,..., n, (II) Subtract the arithmetic mean value of the sensor measurement data obtained in step (I) from the subsequent sampling value x h of the positioning system, (III) Set the system required error. When the difference value between the arithmetic mean value of the sensor measurement data and the subsequent sampling value of the positioning system is smaller than the system required error, it is shown that the unmanned ship position data collected by the multi-sensor positioning system has consistency and is reliable data. When the difference value is larger than the system required error, dispersion weighted compensation is performed on the sampling data, and thus it is necessary to meet the sampling requirement of the data sample by basic particle filtering. (IV) The arithmetic mean value [Number] As the offset-free estimated value, when the unmanned ship multi-sensor positioning system measures the unmanned ship navigation trajectory at different positions in the same space, the information measurement variance of the i-th sensor system [Number] (V) According to the dataset recorded by the unmanned ship multi-sensor positioning system after performing m measurements on the unmanned ship, the j-th measurement data of the i-th sensor is x ij Let it be, and x ij Replace the sensor measurement information x i in step (IV) with this, thereby obtaining the dataset information variance obtained by multiple measurements [Number] In the formula, i = 1, 2,..., n, j = 1, 2,..., m, (VI) Based on the position information dataset variance collected by the multi-sensor calculated in step (V), the fusion weight k i of the inconsistent fault data in step (III) is defined. The calculation of the k i is as follows, [Number] (VII) Based on the obtained fusion weight k i , perform variance weighted fusion on the inconsistent fault data [Number] to obtain sensor measurement data [Number]

[0060] By performing a consistency check on the unmanned ship position information collected by the multi-sensor positioning system and performing weighted compensation on the fault data, the fault tolerance performance of the data fusion algorithm is improved, and the reliability of the particle set samples sampled by the basic particle filtering algorithm is ensured.

[0061] As shown in Figure 5, based on the basic particle filtering algorithm, the method for performing filtering processing on the multi-sensor data after preprocessing, inspection, and compensation respectively adopts the following steps: (i) Establish a basic particle filtering model for the unmanned ship multi-sensor positioning system, substitute the unmanned ship position information after consistency check and variance weighting into the system model, and describe the state and measurement models of the sensor system by the following equations

Number

Number

Number

Number

[0062] Particle propagation samples the unmanned ship multi-sensor positioning system state transition model [Number] and then generates a new particle state X k . Here, X k-1 is the particle state after the previous resampling, and Z k is the detection data of the sensor system. The measurement data of the unmanned ship multi-sensor positioning system is filtered using the basic particle filtering algorithm to realize data augmentation, reduce the interference of the measurement data by environmental noise, ensure the processing accuracy of the sampling sample data set, and further improve the reliability of the unmanned ship multi-sensor measurement data.

[0063] As shown in FIGS. 6 and 7, the method for constructing the Gaussian mixture model adopts the following steps (a) Extract the multi-sensor data set processed by the basic particle filtering algorithm and the Sigma point set [Number] ​ The set of Sigma points after unscented transformation, where n a is the dimension of the Sigma points, λ is the scale parameter, and (b) Fuse the latest measurement information with the obtained Sigma sampling point set and the system state

Number

Number

Number

Number

Number

Number

Number

[0064] In the importance sampling process of the novel threshold hierarchical particle filtering algorithm, the current measurement information is incorporated into the particle set proposal distribution, making the proposal distribution closer to the actual posterior probability density and improving the algorithm estimation performance. As can be seen from Figure 6, the constructed weighted Gaussian mixture distribution function satisfies a continuous multimodal normal distribution at each time, and it is shown that the weighted point set mixture Gaussian distribution composed of similar component units is close to the posterior probability density function. Also, it is shown that the function peak distribution is concentrated and the particle samples effectively represent the probability distribution characteristics after resampling.

[0065] As shown in Figures 8 and 9, the specific steps of the method for setting the hierarchical sampling ratio capacity related to the adaptive threshold and the confidence factor, further performing fusion filtering on the unmanned ship navigation trajectory positioning data, and outputting the unmanned ship navigation trajectory positioning information are as follows: a) After selecting the importance sampling process, the particle X with the maximum weight in the discrete particle sample set c is used as the clustering center, and the Martensite distance D i between it and other particles i is calculated, and the D i is shown as follows,[ [Number] is the probability density of particle i, and S is the covariance matrix, b) Calculate the number of effective particle samples N e in the clustering unit, and the N eThe calculation is as follows:

Number

Number

Number

Number

[0066] h) Based on the hierarchical theory, each probability density function is a continuous probability density function of p(x)

Number

[0067] i) The ratio capacities of the number of particles in each l a 、l b 、l c layers are set to N / 4, N / 3, and N / 3.

[0068] j) Substitute the confidence factor into the weight optimization combination for calculation.

[0069] k) For particles with weights in the l b 、l c layers that are smaller than the average value

Number

Number

Number

[0070] l) Obtain the multi-sensor data fusion sampling result obtained in step (k).

[0071] m) Output the data fusion sampling result obtained in step (l) from the blind node installed on the unmanned ship in the form of a log file.

[0072] n) By networking the PC - side coordinator node arranged on the riverbed with the unmanned ship blind node, the position information of the unmanned ship output by the blind node in step (u) is obtained in real - time, thereby realizing the positioning of the unmanned ship navigation track.

[0073] Hierarchize the sampling samples of the continuous probability density function of the constructed weighted point set, and set the ratio capacity of each sampling layer to ensure that the allocation of the number of sampling particles in the hierarchy is reasonable, and then a Sample the layer set, and b and c perform an optimization combination on the weights of the particles in the layer set to increase its probability mass, and also associate the measurement data reliability factor collected by the unmanned ship multi - sensor positioning system with the hierarchical sampling in the multi - sensor data algorithm. When performing data fusion on the unmanned ship multi - sensor measurement data using the new threshold hierarchical particle filtering algorithm, preferentially sample and fuse the sensor measurement values with large reliability factors to increase the reference value of the entire information sample. As can be seen from Figure 8, at time k = 30, the sampling points are concentrated at the extreme value of the continuous probability density function. By substituting the sensor data reliability factor into the hierarchical sampling for calculation, when resampling the continuous probability density function, it is ensured that the sampling points are concentrated at the places with larger probability density function values, increasing the reference value of the entire information sample, and further ensuring the fusion positioning accuracy of the measurement data by the new threshold hierarchical particle filtering algorithm.

[0074] To verify the optimized performance of the new threshold hierarchical particle filtering algorithm of the present invention, in this specification, 30 independent tests are performed on the computer simulation test system model for the root mean square error RMSE and the standard deviation Std, and the test results are compared with the Extended Kalman Filter (EKF) [1] and the Unscented Kalman Filter (UKF) [2], Unscented Particle Filter (UPF) [3] and Basic Particle Filter BPF [4] By comparing with this, the effectiveness of the improved step of the filtering algorithm is verified. The simulation time step size k = 50 for the five algorithms, the time interval Δt = 1, and the parameters of the comparison target algorithms are taken from the corresponding reference documents.

[0075] Attached computer simulation test model: [Number] References are attached: [1] He Junyi, Li Nannan, An Wei Peng. Research on Improved Extended Kalman Filtering Algorithm [J]. Prediction and Control Technology, 2018, 37(12): 102 - 106. [2] Li Xingjia, Li Jianfen, Zhu Min, etc. Research on Positioning Fusion and Detection Algorithm Based on Unscented Kalman Filtering [J]. Automotive Engineering, 2021, 43(6): 825 - 832. [3] Wu Bin, Tian Qing. Indoor Moving Target Positioning Algorithm of Improved Unscented Particle Filtering [J]. Sensors and Microsystems, 2021, 40(3): 153 - 156, 160. [4] Gao Yi, Mao Yanhui, Yang Yi. Artificial Fish School Particle Filtering Algorithm [J]. Modern Electronics Technique, 2021, 44(16): 170 - 174.

[0076] As can be seen from Table 1, Figure 10(a), and Figure 10(b), whether it is the average value or the maximum value of RMSE, in the specification, TLPF is smaller than the other four algorithms in both cases, indicating that the filtering accuracy of the novel threshold hierarchical particle filtering algorithm in this patent is the highest. This is mainly because in the specification, an adaptive threshold is constructed in the Gaussian mixture to cluster discrete particles, and a weight optimization combination is performed on the particles in the inferior layer to improve the diversity of particle samples. From the Std average value and the maximum value, it can be seen that compared with the other four algorithms, the value of TLPF is also the smallest, indicating that the filtering stability of TLPF is the best. This is mainly due to the selection of the importance function in the importance sampling process. The latest measurement data is fused into the importance function through the unscented transformation to improve the reliability of particle samples, and a re-hierarchical sampling calculation is performed on the particles in the inferior layer in the resampling process, substituting more particle samples into the sampling process, thereby ensuring the diversity of the particle sample set.

[0077] To verify the optimized performance of the multi-sensor data fusion algorithm of the present invention, a comparative positioning test of the multi-sensor system algorithm based on an unmanned ship platform was carried out. 100 sets of positioning data collected by the multi-sensor system in the test river channel were selected respectively. The root mean square error RMSE and the standard deviation Std were used as the performance indicators of the test results, and the test results were compared with the extended Kalman filtering, unscented Kalman filtering, unscented particle filtering, and basic particle filtering. Figure 4 shows the results of performing filtering fusion on the unmanned ship navigation positioning trajectory data with each algorithm.

[0078] As can be seen from Table 2, Figure 11, and Figure 12, when the unmanned ship sails in the test river channel, due to the dense trees and narrow and thin river channel, it has a certain environmental perturbation impact on the sensor signal transmission. The filtering positioning results of each data fusion algorithm deviate significantly from the actual path of the unmanned ship's navigation in some places. Moreover, at the curve of the path, large jumps are likely to occur in the positioning results, gradually deviating from the actual trajectory of the unmanned ship's navigation, reducing the reliability of the positioning results. As time goes by, the error of the positioning results is becoming larger and larger. However, the positioning accuracy of the multi-sensor data fusion algorithm of the present invention is higher than that of the fusion positioning accuracy of any algorithm. This is mainly because in the present invention, first, through consistency inspection, fault inspection is performed on the unmanned ship multi-sensor measurement data, and weighted compensation processing is performed on the inconsistent data to improve its reliability. When the basic particle filtering performs data augmentation on the data samples, it has more reliable sampling samples, improving the processing accuracy and fault tolerance performance of the data fusion algorithm. Next, through reliability inspection, reliability inspection is performed on the data collected by the unmanned ship multi-sensor positioning system, and corresponding positioning data reliability factors are assigned according to the magnitude of the reliability distance, and it is associated with the hierarchical sampling operation steps of the subsequent sensor data fusion algorithm. Finally, by using the new threshold hierarchical particle filtering algorithm, a continuous probability density function of Gaussian mixture is constructed, and the clustering integration efficiency is improved by setting an adaptive threshold, improving the real-time performance of the data fusion algorithm for positioning data processing, and associating the reliability factor with the hierarchical sampling weight calculation, preferentially sampling the particle samples with high reliability factors, and improving the positioning accuracy of the data fusion algorithm. The multi-sensor data fusion algorithm of the present invention reduces the average positioning error by 47% compared with the other four data fusion algorithms, fully ensuring the positioning accuracy for the unmanned ship navigation trajectory.

[0079]

Table 1

[0080]

Table 2

[0081] The above embodiments are only used to explain the technical solutions of the present invention and are not limited thereto. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art can modify or equivalently replace the specific embodiments of the present invention, and should understand that any modification or equivalent replacement without departing from the spirit and scope of the present invention should be included within the protection scope of the claims of the present invention.

Claims

1. A positioning method for an unmanned ship based on multi-sensor data fusion, comprising: Step (1) of preprocessing the unmanned ship positioning data collected by the unmanned ship multi-sensor positioning system, the specific content and method steps of which are as follows: (A) Based on the longitude and latitude information of the unmanned ship navigation track collected by the positioning system sensor, unify the time and space reference therefor; (B) According to the river channel along which the unmanned ship sails, construct a communication environment that satisfies the effective connection of the multi-nodes of the positioning system, so that the blind node located on the unmanned ship is within a network composed of four signal receiving nodes with known transmission signal strength and position coordinates, thereby calculating and collecting the position coordinates when the unmanned ship sails through the river channel; (C) Step (1) of converting and unifying the multi-sensor positioning system coordinates based on the Gauss-Kruger projection principle; Step (2) of performing reliability determination and reliability factor assignment based on the data collected by the multi-sensor; Step (3) of performing a consistency check on the multi-sensor data, and performing weighted compensation processing on the inconsistent fault data collected by the positioning system; Step (4) of performing filtering processing on the preprocessed, inspected, and compensated multi-sensor data respectively based on the basic particle filtering algorithm; Construct a Gaussian mixture model, and design a new threshold hierarchical particle filtering algorithm by setting an adaptive threshold and a hierarchical sampling ratio capacity related to the reliability factor, and further perform fusion filtering on the unmanned ship navigation track positioning data to output the unmanned ship navigation track positioning information. A positioning method for an unmanned ship based on multi-sensor data fusion, characterized in that it comprises the above steps.

2. The specific content and method steps of the above step (2) of performing reliability determination and reliability factor assignment based on the data collected by the multi-sensor are as follows: 1) Based on the measurement data collected by the positioning system, construct the multi-sensor measurement model p ｉ (x), and calculate the trust distance d ｉ between sensor data, and calculate the p ｉ (x) and d ｉ respectively according to the following equations: 【Number 85】 In the formula, x ｉ is the measured value of the i-th sensor, μ is the true value of the measurement feature, and θ ｉ is the measurement accuracy of the i-th sensor information, and σ ｉ is the measurement error of the i-th sensor information, 【Number 86】 where each of them is a measurement value collected by the multi-sensor at time i; 【Number 87】 is the measurement dispersion average value, Z is a random variable following the standard normal distribution, and i = 1, 2,..., n; 2) Rewrite d obtained in step 1) according to the Gaussian probability model into a metric p in the probabilistic sense between sensor measurement data ｉ and set the sensor support confidence level, thereby judging the reliability of different sensor information, and calculate the p ｒ (Z ｉ ) as follows, and ｒ (Z ｉ ) is calculated by the following formula 【Number 88】 In the formula, Z ｉ is the measurement data of the multi-sensor, ε is the confidence level, K is the variable coefficient of the sampling sample probability interval, 3) According to the sensor reliability distance, judge the magnitude of the reliability of the position information collected by the multi-sensor, and divide the positioning data into different reliability intervals based on the magnitude of the information reliability; 4) Dynamic support degree factor β of measurement information collected by the multi-sensor at time k ｉ (k) and the norm equation of the Gaussian probability measurement model p ｉ (k) are constructed, and the β ｉ (k) is calculated by the following formula, 【Number 89】 is the Frobenius norm, k = 1, 2,..., T, and i = 1, 2,..., n. 5) The dynamic support degree factor β of the multi-sensor obtained in step 4) ｉ (k), based on which, the measurement error w of the system ｉ is calculated, and the w ｉ is calculated by the following formula w ｉ = z ｉ - Aβ ｉ (k) In the formula, A is a state transition matrix, and z ｉ is a measurement value of the unmanned ship multi-sensor positioning system, where i = 1, 2,..., n, 6) Variance of the measurement error w of the system ｉ of 【Number 90】 confidence factor corresponding to measurement data using 【Number 91】 A confidence factor of sensor measurement data, where i = 1, 2, …, n, the unmanned ship positioning method based on multi-sensor data fusion according to claim 1, characterized in that.

3. By performing a consistency check on multi-sensor data, the specific content and method steps of step (3) for performing weighted compensation processing on inconsistent failure data collected by the positioning system are as follows: (I) Arithmetically average the sensor measurement data obtained by the unmanned ship positioning system to obtain an arithmetic mean value 【Number 92】 In the formula, x ｉ is sensor measurement information, where i = 1, 2,..., n, (II) Subtract the arithmetic mean value of the sensor measurement data obtained in step (I) from the subsequent sampling value x of the positioning system ｈ and (III) Set the system required error. When the difference value between the arithmetic mean value of the sensor measurement data and the subsequent sampling value of the positioning system is smaller than the system required error, it is determined that the unmanned ship position data collected by the multi-sensor positioning system has consistency and is reliable data. When the difference value is larger than the system required error, weighted compensation for the sampling data is performed, and thus it is necessary to satisfy the sampling requirement of the data sample by basic particle filtering. (IV) The arithmetic mean value of multi-sensor measurement data 【Number 93】 As an offset-free estimated value of, when the unmanned ship multi-sensor positioning system measures the unmanned ship navigation trajectory at different positions in the same space, the information measurement variance of the i-th sensor system 【Number 94】 (V)According to the data set recorded by the unmanned ship multi-sensor positioning system through m measurements on the unmanned ship, the measurement data of the j-th measurement of the i-th sensor is x ｉｊ Let it be, and x ｉｊ Replace the sensor measurement information x ｉ in step (IV) with this, thereby obtaining the data set information dispersion obtained by multiple measurements 【Number 95】 In the formula, i = 1, 2, …, n and j = 1, 2, …, m. (VI) Based on the dispersion of the position information data set collected by the multi-sensor calculated in step (V), the fusion weight k of the mismatch fault data in step (III) ｉ is defined, and the said k ｉ is calculated by the following formula, 【Number 96】 (VII) The obtained fusion weight k ｉ Based on this, the inconsistent data 【Number 97】 Perform weighted compensation on, thereby obtaining sensor measurement data that satisfies the basic particle filtering processing requirement 【Number 98】 Calculate using the following formula, the unmanned ship positioning method based on multi-sensor data fusion according to claim 1, characterized in that. 【Number 99】

4. Based on the basic particle filtering algorithm, the specific content and method steps of step (4) for performing filtering processing on the pre-processed, inspected, and compensated multi-sensor data are as follows: (i) Establish the basic particle filtering model of the unmanned ship multi-sensor positioning system, substitute the unmanned ship position information after consistency check and variance weighting into the system model, and describe the state and measurement model of the sensor system using the following formula 【Number 100】 In the formula, x ｋ is the predicted position value of the sensor system at time k, and x ｈ is the subsequent sampling value of the sensor. 【Number 101】 is the sensor measurement value after distributed weighting, z ｋ is the measured value of the unmanned ship position at time k, λ ｋ is the estimated noise, v ｋ is the measurement noise (ii) Initialize the particle set sample and randomly sample from the prior density p(x ０ ) to initialize the particle set 【Number 102】 (iii) Randomly extract N particle samples from the importance density function. (iv) The weight of the sampling particle 【Number 103】 It is the weight of the sampling particle, where i = 1, 2, …, N. (v) Normalize the importance weight. (vi) The number of effective particles N in the elementary particle filtering algorithm ｅｆｆ is calculated, and compared with the threshold value N ｔｈ and when N ｅｆｆ < N ｔｈ resampling is performed, and the said N ｅｆｆ is calculated by the following formula, 【Number 104】 Output the (vii) state, and calculate the partially estimated measurement information collected by the unmanned ship multi-sensor positioning system that has been filtered 【Number 105】 respectively according to the following formulas, and the unmanned ship positioning method based on multi-sensor data fusion according to claim 1, characterized in that. 【Number 106】

5. The specific content and method steps of the step (5) of constructing the Gaussian mixture model are as follows: (a) Extract the multi-sensor data set processed by the basic particle filtering algorithm, and the Sigma point set [Number 107] The Sigma point set after unscented transformation, where n ａ is the dimension of the Sigma points, and λ is the scale parameter, (b) Fuse the latest measurement information with the obtained Sigma sampling point set, and the system state 【Number 108】 In the formula, k ｋ is the Kalman gain, and z ｋ is the measurement information, 【Number 109】 is the covariance measured by the weighted Sigma point set. (c) The multi-sensor positioning system state obtained in step (b) 【Number 110】 and sample from the proposed distribution constructed. (d) According to the Gaussian mixture model, the posterior probability density function with a time step size of k 【Number 111】 is the i-th component in the Gaussian mixture model, C(k) is the number of component units of the discrete sample, and ξ is the discrete point component weight. (e) The discrete sampling points sampled in step (c) and their corresponding weights 【Number 112】 are fused into the Gaussian mixture component unit in step (d), and the continuous probability density function constructed 【Number 113】 In the formula, p(k) is the covariance of the discrete particle filtering distribution. 【Number 114】 is the mean value of the discrete particle filtering distribution, h is the normalization constant, and n ｘ is the particle distribution dimension, (f) Adopt clustering analysis to perform an integration process on the Gaussian mixture similar units in the continuous posterior probability density function in step (e). 【Number 115】 The unmanned ship positioning method based on multi-sensor data fusion according to claim 1, characterized in that.

6. The specific content and method steps of the step (5) of setting the adaptation threshold are as follows: After selecting the importance sampling process, the particle X with the maximum weight in the discrete particle sample set ｃ is used as the clustering center, and the martensite distance D ｉ therebetween and other particles i is calculated, and the D ｉ is shown below. 【Number 116】 is the probability density of particle i, and S is the covariance matrix. (h) The number N of effective particle samples in the clustering unit ｅ Calculate, and for the said N ｅ Calculate using the following formula: 【Number 117】 is the particle probability density covariance. (i) Construct the threshold T, and the T is shown below. 【Number 118】 In the formula, T ０ is the threshold initial value, k ｅ is the ratio coefficient, and R is the number of classification times (j) The number N of effective particle samples obtained in step (h) ｅ is substituted into the threshold value T to construct the adaptive threshold value T ｃ , and the T ｃ is shown below 【Number 119】 (k) D ｉ and the adaptation threshold T ｃ are compared, and if D ｉ is smaller than T ｃ the particle is placed in the component unit related to its probability mass, and if D ｉ is larger than T ｃ this particle is skipped and other particles are clustered (m) Select the particle with the maximum weight from other particle samples as the clustering center, and repeat step (k) until clustering is completed. (n) According to the component units after clustering, the continuous probability density function of the constructed particle set. 【Number 120】 In the formula, β ｉ is the probability mass of the similar component unit component i, γ ｉ is the average value of the component unit component i, p ｉ is the covariance of the component unit component i, where i = 1, 2, …, n. The unmanned ship positioning method based on multi-sensor data fusion according to claim 1, characterized by this.

7. Relate the confidence factor to the hierarchical sampling ratio capacity, and further perform fusion filtering on the unmanned ship navigation trajectory positioning data. The specific content and method steps of the step (5) for outputting the unmanned ship navigation trajectory positioning information are as follows: (o) Based on the hierarchical theory, a continuous probability density function whose respective probability density function is p(x) 【Number 121】 Divide it into l layers, and divide the grouped layers into one set of weight-dominant layers and two sets of inferior layers according to the magnitude of the probability mass, and respectively l ａ 、l ｂ 、l ｃ Define it as, (p) l for each ａ , l ｂ , l ｃ The ratio capacities of the number of particles in the l layers are N / 4, N / 3, and N / 3, (q) Substitute the confidence factor into the weighted optimization combination for calculation, (r) l ｂ , l ｃ The weight in the layer is the average value 【Number 122】 Perform weighted optimization combination on particles smaller than, and obtain the weight of the optimized particles, 【Number 123】 And perform hierarchical sampling on the sample data, 【Number 124】 (s) Obtain the multi-sensor data fusion sampling result obtained in step (r), (t) Fuse the data obtained in step (s) with the sampling result, and output it in the form of a log file from the blind node installed on the unmanned ship, (u) The PC-side coordinator node arranged on the riverbed is networked with the unmanned ship blind node to obtain the unmanned ship position information output by the blind node in real time in step (u), thereby realizing the positioning of the unmanned ship navigation trajectory. The unmanned ship positioning method based on multi-sensor data fusion according to claim 1, characterized in that.

Images