A wide-frequency scattering source cooperative positioning system based on an underwater sensor network
By using temperature, salinity, and depth interpolation, Thorp attenuation model, and joint signal modeling in the spatial and frequency domains of a distributed underwater sensor network, the problem of insufficient accuracy caused by sound speed variation, multipath effect, and node movement in traditional underwater positioning technology is solved, achieving high-precision and low-cost broadband scattering source positioning.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- SOUTH CHINA UNIV OF TECH
- Filing Date
- 2026-05-06
- Publication Date
- 2026-07-07
AI Technical Summary
Traditional underwater positioning technology suffers from insufficient positioning accuracy due to factors such as non-uniform sound speed, multipath effect, and broadband signal attenuation, and node movement leads to increased errors.
A distributed underwater sensor network is adopted, and sound velocity variation is corrected by temperature, salinity and depth interpolation, signal amplitude is compensated by Thorp attenuation model, node position is dynamically calibrated, multipath components are eliminated, and collaborative localization is achieved by using joint signal modeling in the spatial and frequency domains and low-rank reconstruction method.
It improves positioning accuracy and precision, reduces system deployment costs, and enables efficient positioning of multiple broadband scattering sources.
Smart Images
Figure CN122131233B_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of underwater sensor network positioning technology, specifically a broadband scattering source cooperative positioning system based on underwater sensor networks. Background Technology
[0002] With the development of marine monitoring technology, underwater scattering source location technology has become one of the core technologies in the field of marine monitoring. Traditional underwater positioning technologies mainly include the Time of Arrival (TOA) positioning method and the Time Difference of Arrival (TDOA) positioning method. These methods can achieve a certain positioning accuracy in ideal environments. However, in actual underwater environments, traditional positioning methods have many shortcomings: the speed of sound underwater is not uniformly distributed and changes with temperature, salinity, and depth. Most traditional positioning methods assume that the speed of sound is uniform, which leads to a large error in the calculation of the direction vector, thus resulting in a decrease in positioning accuracy.
[0003] Meanwhile, underwater acoustic signals suffer from severe multipath effects, with reflected signals interfering with direct signals. Traditional positioning methods cannot effectively eliminate multipath components, leading to significant deviations in positioning results. Furthermore, underwater acoustic signals are broadband signals, with different frequencies exhibiting varying degrees of attenuation. Traditional positioning methods do not compensate for this broadband attenuation, resulting in signal amplitude distortion and affecting positioning accuracy. Moreover, underwater sensor nodes move with ocean currents. Most traditional positioning methods assume the nodes' positions are fixed and do not dynamically calibrate them, causing positioning errors to increase over time. Summary of the Invention
[0004] The purpose of this invention is to propose a broadband scattering source cooperative localization system based on an underwater sensor network, comprising:
[0005] Distributed underwater sensor node groups are deployed in the underwater monitoring area to collaboratively collect underwater acoustic signals from broadband scattering sources.
[0006] The parameter initialization configuration module configures the system's operating parameters and completes parameter initialization before positioning.
[0007] The spatial and frequency domain joint signal modeling module transforms the received signals from the distributed underwater sensor node group and establishes a joint signal model of spatial and frequency domain distribution.
[0008] The collaborative reconstruction basis construction module constructs a reconstruction basis matrix adapted to multi-node collaboration based on initialization parameters and joint signal model.
[0009] The sampling signal preprocessing module performs vectorization processing on the sampling matrix composed of the received signals collected by each node in row-major order to obtain a standardized input vector.
[0010] The low-rank constraint optimization solution module constructs a low-rank matrix reconstruction optimization problem based on the reconstructed basis and input vector, and applies a nuclear norm low-rank constraint to the joint space-frequency distribution matrix to complete the problem solution.
[0011] The scattering source parameter calculation module calculates the location parameters of the broadband scattering source based on the obtained spatial-frequency joint distribution density matrix.
[0012] The technical solution of the present invention brings at least the following beneficial effects:
[0013] By using temperature, salinity, and depth interpolation and sound velocity profile correction, the influence of sound velocity variations on the direction vector can be effectively corrected, significantly improving positioning accuracy. Amplitude compensation for signals of different frequencies using the Thorp attenuation model effectively corrects attenuation differences in broadband signals, improving signal quality and thus positioning accuracy. Dynamic node position calibration, with the calibration cycle dynamically adjusted according to ocean current velocity, effectively tracks node movement, ensuring real-time node position and avoiding positioning errors caused by node movement. Multipath component removal effectively filters out reflected interference signals, retaining only direct signals, significantly improving positioning accuracy. A low-rank reconstruction method combining spatial and frequency domains enables simultaneous positioning of multiple broadband scattering sources without requiring strict time synchronization, reducing system deployment costs. Attached Figure Description
[0014] Figure 1 This is a flowchart of the method of the present invention. Detailed Implementation
[0015] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.
[0016] Please see Figure 1 This application provides a broadband scattering source cooperative localization system based on an underwater sensor network, comprising:
[0017] Distributed underwater sensor node groups are deployed in the underwater monitoring area to collaboratively collect underwater acoustic signals from broadband scattering sources.
[0018] The parameter initialization configuration module configures the system's operating parameters and completes parameter initialization before positioning.
[0019] The spatial and frequency domain joint signal modeling module transforms the received signals from the distributed underwater sensor node group and establishes a joint signal model of spatial and frequency domain distribution.
[0020] The collaborative reconstruction basis construction module constructs a reconstruction basis matrix adapted to multi-node collaboration based on initialization parameters and joint signal model.
[0021] The sampling signal preprocessing module performs vectorization processing on the sampling matrix composed of the received signals collected by each node in row-major order to obtain a standardized input vector.
[0022] The low-rank constraint optimization solution module constructs a low-rank matrix reconstruction optimization problem based on the reconstructed basis and input vector, and applies a nuclear norm low-rank constraint to the joint space-frequency distribution matrix to complete the problem solution.
[0023] The scattering source parameter calculation module calculates the location parameters of the broadband scattering source based on the obtained spatial-frequency joint distribution density matrix.
[0024] It should be noted that the distributed node group is responsible for distributed signal acquisition, the parameter initialization module completes the preparation of various parameters before positioning, the spatial frequency domain modeling module establishes a signal model adapted to the underwater environment, the reconstruction base module constructs a reconstruction base adapted to multi-node collaboration, the preprocessing module completes the standardization of input data, the low-rank solution module implements the core positioning optimization calculation, and the parameter solution module completes the final positioning parameter calculation. All modules cooperate with each other to achieve collaborative positioning of underwater broadband scattering sources.
[0025] As an optional embodiment, the distributed underwater sensor node group includes a node deployment unit, a self-organizing networking unit, a dynamic position calibration unit, and an underwater acoustic clock synchronization unit;
[0026] The node deployment unit reads the pre-stored historical ocean current distribution data and topographic data of the monitoring area, and deploys the underwater sensor nodes in the target sea area. The distance between the nodes is set to be less than the maximum transmission distance of the underwater acoustic communication module used. The union of the monitoring ranges of all nodes covers the entire target sea area.
[0027] The self-organizing networking unit broadcasts a neighbor discovery request after the node is powered on. After receiving the request, the neighbor node obtains the distance to the neighbor node through TOF ranging, replies with a response, builds a communication route, and forms a self-organizing network. Each underwater sensor node integrates an underwater acoustic transducer, a signal acquisition unit, and a short-range underwater acoustic communication unit. The communication unit adjusts the transmission power according to the distance to the neighbor node measured during the neighbor discovery process. The greater the distance, the greater the transmission power.
[0028] The dynamic position calibration unit acquires bidirectional sonar ranging data obtained through two-way sonar interaction between nodes, reads node depth data collected by the depth sensors on each node, calculates the real-time position of the node based on the ranging data and depth data, updates the node's position parameters to the calculated real-time position, and adjusts the calibration cycle according to the real-time ocean current velocity of the monitored sea area; the higher the current velocity, the shorter the calibration cycle.
[0029] The underwater acoustic clock synchronization unit works as follows: Node A sends a data packet carrying a local timestamp to Node B. Node B records its local time when it receives the data packet. Node B then replies to Node A with a data packet carrying the received time and its local timestamp. Node A records its local time when it receives the data packet. Based on these four timestamps, the clock deviation between the nodes is calculated, and clock deviation calibration between the distributed nodes is completed.
[0030] It should be noted that, firstly, the historical ocean current distribution data and topographic data of the monitoring area are read. This historical data comes from publicly available data in the field of marine monitoring and can be used to guide the deployment of nodes.
[0031] Underwater sensor nodes are deployed in the target sea area, with the distance between nodes set to be less than the maximum transmission distance of the underwater acoustic communication module used, to ensure normal communication between the nodes. Simultaneously, the monitoring ranges of all nodes are combined to cover the entire target sea area, avoiding monitoring blind spots.
[0032] During the self-organizing network phase, after powering on, each node first broadcasts a neighbor discovery request. Upon receiving the request, neighboring nodes calculate the distance between the two nodes using Time-of-Flight (TOF) ranging, which measures the signal propagation time. They then reply to the requesting node, thus establishing a communication route and forming a self-organizing network. Each underwater sensor node integrates an underwater acoustic transducer, a signal acquisition unit, and a short-range underwater acoustic communication unit. The communication unit adjusts its transmission power based on the distance to neighboring nodes measured during discovery; the greater the distance, the higher the transmission power. This adjustment is based on the path loss characteristics of underwater acoustic communication to ensure communication reliability.
[0033] During the dynamic position calibration phase, bidirectional sonar ranging data obtained through two-way sonar interaction between nodes is acquired. This involves node A sending an acoustic signal to node B, node B recording the reception time, and then node B sending an acoustic signal back to node A, with node A recording the reception time. The propagation time in both directions eliminates the influence of clock deviation, resulting in more accurate distance data. Simultaneously, depth data collected by the depth sensors on each node is read. Based on this ranging and depth data, the real-time position of the node is calculated, and the node's position parameters are then updated to reflect the calculated real-time position. The calibration cycle is set based on the real-time ocean current velocity in the monitored sea area. Specifically, it adopts the relationship t=600 / v, where v is the ocean current velocity and t is the calibration cycle. This relationship is derived from the physical laws of underwater node movement. The greater the current velocity, the faster the node moves, so a shorter calibration cycle is needed to ensure the accuracy of the node's position. For example, when the current velocity is 1 m / s, the calibration cycle is 600 seconds, and when the current velocity is 2 m / s, the calibration cycle is 300 seconds. This ensures that the node's movement distance between two calibrations will not exceed the positioning accuracy requirements.
[0034] During the time synchronization phase, the TPSN time synchronization algorithm is used. Specifically, node A sends a data packet carrying its local timestamp to node B. When node B receives the data packet, it records its local time. Then, node B replies to node A, carrying the received time and its local timestamp. When node A receives the data packet, it also records its local time. Finally, node B's received time is subtracted from node A's sent time, and the difference is divided by 2 to obtain the clock skew between the nodes. This completes the clock skew calibration between distributed nodes and eliminates the time difference between nodes.
[0035] As an optional embodiment, the parameter initialization configuration module includes an array parameter configuration unit, a search range configuration unit, a discrete precision configuration unit, and a sound velocity profile adaptation unit.
[0036] The array parameter configuration unit reads the position parameters of each underwater sensor node after dynamic calibration and completes the parameter configuration of the distributed node array.
[0037] The search range configuration unit configures the target angle search range according to the lateral range of the monitored sea area, and configures the system's operating wideband range according to the frequency range of the radiation signals from the pre-stored scattering sources.
[0038] The discrete precision configuration unit configures the step size parameter of spatial frequency domain discretization according to the preset maximum positioning error. The smaller the preset maximum positioning error, the smaller the corresponding step size. The preset maximum positioning error ranges from 0.01 degrees to 1 degree.
[0039] The sound velocity profile adaptation unit loads real-time temperature, salinity, and depth data collected by the temperature, salinity, and depth sensors built into each underwater sensor node. It obtains the temperature, salinity, and depth distribution of the entire monitored sea area through linear interpolation, calculates the sound velocity profile data of the entire sea area based on the empirical formula for sound velocity, and corrects the propagation delay parameter of the direction vector based on the sound velocity profile, providing sound propagation parameters for the calculation of the direction vector.
[0040] It should be noted that the position parameters of each underwater sensor node are read after dynamic calibration to complete the parameter configuration of the distributed node array.
[0041] During the search range configuration phase, the search range for the target angle is calculated based on the lateral extent of the monitored sea area to ensure that all possible scattering source locations are covered without omission. The system's operating bandwidth is configured based on the frequency range of the radiation signals from the scattering sources, which are pre-stored parameters input by the user before the mission starts, ensuring that all scattering source signals can be acquired.
[0042] During the discrete precision configuration phase, the preset maximum positioning error is the upper limit of the positioning error configured by the user before the task starts. The step size setting follows the Nyquist sampling theorem for discrete sampling; the smaller the step size, the higher the sampling accuracy. Therefore, when the user needs higher positioning accuracy, a smaller step size is set to ensure that the sampled signal can completely reconstruct the original distribution information. The preset maximum positioning error ranges from 0.01 degrees to 1 degree, covering the accuracy requirements of all conventional positioning tasks.
[0043] During the sound velocity profile adaptation phase, each underwater sensor node integrates a standard temperature, salinity, and depth (TDM) sensor, enabling real-time acquisition of temperature, salinity, and depth data at its location. Since underwater sound velocity varies with location, linear interpolation is used to expand the single-point TDM data collected by each node into a TDM distribution for the entire monitored sea area. Linear interpolation is a commonly used interpolation method in numerical computation, effectively expanding discrete single-point data into continuous distributed data. Then, based on empirical formulas for sound velocity, the sound velocity profile data for the entire sea area is calculated. The speed of sound was calculated using the empirical formula proposed by Wilson in 1960. The specific calculation process is as follows: First, the temperature value is multiplied by 4.6, then the square of the temperature value multiplied by 0.055 is subtracted, and the cube of the temperature value multiplied by 0.00029 is added. Next, a salinity correction term is calculated by subtracting the temperature value multiplied by 0.01 from 1.34, multiplying by the salinity value minus 35, and then adding a depth correction term by multiplying the depth value by 0.016. Finally, the base sound speed value of 1449.2 is added to obtain the speed of sound at the current location. This formula is applicable to temperatures from -2 degrees Celsius to 30 degrees Celsius, salinity from 0 to 40, and depths from 0 to 8000 meters, completely covering the monitoring area of this application. Finally, based on the obtained sound speed profile, the propagation delay parameter of the direction vector is corrected to correct the error in the direction vector caused by the difference in sound speed at different locations.
[0044] As an optional embodiment, the spatial-frequency domain joint signal modeling module includes a frequency domain transformation unit, a wideband attenuation compensation unit, and a time domain reconstruction unit;
[0045] The frequency domain transformation unit performs frame processing on the received time domain signal of each underwater sensor node by setting the frame length according to the signal length, setting a frame overlap rate of 50%, performing windowing processing on the signal using a Hamming window, performing amplitude compensation on the windowed signal, setting an FFT length of the same as the frame length, and performing a fast Fourier transform to convert the time domain signal into a frequency domain signal.
[0046] The wideband attenuation compensation unit calculates the attenuation coefficient corresponding to different frequencies based on the Thorp attenuation model, and multiplies the amplitude of the frequency domain signal by the reciprocal of the corresponding attenuation coefficient to complete the amplitude compensation.
[0047] The time-domain reconstruction unit performs an inverse Fourier transform on the compensated frequency-domain signal and restores the framed signal to a complete time-domain signal through the overlapping addition method.
[0048] As an optional embodiment, after processing by the frequency domain transformation unit, the frequency domain signal satisfies the following: for each frequency point, the frequency domain signal is equal to the sum of the spatial integration results of all scattering sources and the environmental noise. The direction vector in the integration term is the multi-node joint direction vector after sound speed profile correction, the integration variable is the target angle, and the integrand is the product of the direction vector and the spatial frequency distribution density of the corresponding scattering source.
[0049] It should be noted that for the received time-domain signal of each underwater sensor node, the frame length is first set according to the signal length, and then a 50% frame overlap rate is set to avoid signal loss caused by framing and ensure signal continuity. Next, a Hamming window is used to window the signal. The Hamming window is a commonly used window function in signal processing, effectively suppressing spectral leakage. Since the Hamming window reduces the signal amplitude, amplitude compensation is performed after windowing to ensure that the signal amplitude is not distorted. Then, an FFT length of the same as the frame length is set, and a Fast Fourier Transform is performed to convert the time-domain signal into a frequency-domain signal.
[0050] In the broadband attenuation compensation stage, the underwater acoustic attenuation model proposed by Thorp in 1967 is used to calculate the attenuation coefficients corresponding to different frequencies. The specific calculation process is as follows: Take the frequency value, first calculate the square of the frequency, then multiply 0.11 by the square of the frequency, divide by 1, add the sum of the squares of the frequencies, add 44 multiplied by the square of the frequency, divide by 4100, add the sum of the squares of the frequencies, add 2.75e-4 multiplied by the square of the frequency, and add 0.003 to obtain the attenuation coefficient value. Then, the attenuation coefficient is converted into an amplitude compensation coefficient. The amplitude of the frequency domain signal is multiplied by the corresponding compensation coefficient to complete the amplitude compensation and correct the attenuation difference at different frequencies.
[0051] In the time-domain reconstruction stage, an inverse Fourier transform is performed on the compensated frequency-domain signal, and then the framed signal is restored to the complete time-domain signal through the overlapping addition method. The overlapping addition method is a commonly used frame reconstruction method in the field of signal processing, which effectively restores the framed signal to the original complete signal.
[0052] As an optional embodiment, after processing by the time-domain reconstruction unit, the time-domain signal satisfies that, for each time point, the time-domain signal is equal to the sum of the space-frequency double integral result of all scattering sources and the time-domain noise. The integral term includes the joint direction vector corrected for sound speed, the Thorp attenuation compensation coefficient, the space-frequency distribution density, and the phase term. The integral variables are the target angle and frequency.
[0053] It should be noted that the calculation rule for the frequency domain signal is as follows: for each frequency point, traverse all scattering sources, integrate over the target angle, and add the result of the integration to the environmental noise to obtain the frequency domain signal at the current frequency point. The direction vector in the integral term is a multi-node joint direction vector after sound speed profile correction, which corrects for the error in the direction vector caused by changes in underwater sound speed.
[0054] As an optional embodiment, the collaborative reconstruction basis construction module includes a dimension discretization unit, a local manifold computation unit, and a cross-node fusion unit;
[0055] The dimension discretization unit reads the initialized discretization step size parameters and performs grid discretization on the angle dimension and frequency dimension to obtain angle sampling points and frequency sampling points;
[0056] The local manifold computation unit calculates the local array manifold vector of each underwater sensor node for each discrete sampling point based on the calibrated node position and sound velocity profile.
[0057] The cross-node fusion unit splices together the local manifold vectors of all nodes to construct a global collaborative reconstruction basis across nodes.
[0058] It should be noted that the calculation rule for the time-domain signal is as follows: for each time point, all scattering sources are traversed, and double integration is performed on the target angle and frequency. The result of the integration is added to the time-domain noise to obtain the time-domain signal at the current time point. The integration term includes a joint direction vector corrected for sound speed, a Thorp attenuation compensation coefficient, spatial frequency distribution density, and a phase term, thereby correcting for the impact of underwater sound speed changes and frequency attenuation on the signal.
[0059] As an optional embodiment, the global collaborative reconstruction basis is based on discrete sampling points as columns. Each column corresponds to a combination of an angle sampling point and a frequency sampling point. The elements of each column are the Kronecker product with the joint direction vector corrected for the speed of sound preceding the time sampling phase vector. The time sampling phase vector is the phase sequence corresponding to the time sampling points after clock synchronization.
[0060] It should be noted that the initial discrete step size parameters are read, and the angle and frequency dimensions are discretized into a grid to obtain the angle sampling points and frequency sampling points.
[0061] In the local manifold calculation stage, for each discrete sampling point, the local array manifold vector of each underwater sensor node is calculated based on the calibrated node position and sound velocity profile.
[0062] In the cross-node fusion stage, the local manifold vectors of all nodes are spliced together to construct a global collaborative reconstruction basis across nodes;
[0063] As an optional embodiment, the low-rank constraint optimization solution module includes a problem transformation unit, an accelerated iterative solution unit, and a convergence verification unit;
[0064] The problem transformation unit constructs an optimization problem by combining the vectorized preprocessed input vector with the collaborative reconstruction basis. The original rank minimization problem is transformed into a least squares optimization problem with a rank upper limit constraint. The rank upper limit is set according to the maximum number of scattering sources preset by the monitoring task. The rank of the space-frequency joint distribution matrix and the number of scattering sources have a 1:1 correspondence.
[0065] The accelerated iterative solution unit employs a Nesterov-accelerated proximal gradient iteration algorithm for the nuclear norm to iteratively solve the optimization problem;
[0066] The convergence verification unit calculates the error for each iteration. When the iteration error is less than the convergence threshold, the iteration is terminated, and the joint spatial frequency distribution density matrix is output. The smaller the preset maximum positioning error, the smaller the corresponding convergence threshold.
[0067] It should be noted that the construction rule for the global collaborative reconstruction basis is as follows: using discrete sampling points as columns, each column corresponds to a combination of angle sampling points and frequency sampling points. The elements of each column are formed by placing the joint direction vector (corrected for sound speed) first, followed by the time-sampled phase vector, and then performing the Kronecker product. The Kronecker product is a standard operation commonly used in matrix operations, combining the features of the two vectors to construct the joint basis vector. The time-sampled phase vector is the phase sequence corresponding to the time sampling points after clock synchronization.
[0068] The vectorized preprocessed input vectors are used to collaboratively reconstruct the basis and construct the optimization problem. The original rank minimization problem is transformed into a least-squares optimization problem with a rank upper bound constraint. The rank upper bound is set according to the maximum number of scattering sources preset by the monitoring task. The rank of the space-frequency joint distribution matrix has a 1:1 correspondence with the number of scattering sources. Each scattering source corresponds to a rank-1 component of the matrix. Therefore, the maximum number of scattering sources is the maximum rank of the matrix. This relationship is derived from the physical meaning of the rank of a low-rank matrix. The preset maximum number of scattering sources is the maximum number of monitoring targets configured by the user before the task starts.
[0069] In the accelerated iterative solution phase, the accelerated optimization algorithm proposed by Nesterov in 1983 is adopted, combined with the proximal gradient iterative algorithm for the nuclear norm, to solve the optimization problem. The low-rank constraint uses the nuclear norm because it is a convex approximation of the rank, effectively constraining the rank of the matrix and facilitating the use of the proximal gradient algorithm. The specific iterative process is as follows: first, the iteration variables are initialized by assigning the initial iteration variables to intermediate variables, and the initial time parameter is set to 1. Then, in each iteration, the gradient of the current step is calculated, the next iteration variable is updated, the next time parameter is calculated, and the next intermediate variable is updated, until the convergence condition is met. The proximal operator for the nuclear norm is a standard operator in the optimization field, used to handle the regularization term of the nuclear norm.
[0070] During the convergence verification phase, the error of each iteration is calculated. When the error of an iteration is less than the convergence threshold, the iteration is terminated, and the joint spatial frequency distribution density matrix is output. The convergence threshold is set based on a preset maximum positioning error. The smaller the preset maximum positioning error, the smaller the corresponding convergence threshold, thus ensuring that the results after iteration meet the positioning accuracy requirements.
[0071] As an optional embodiment, the scattering source parameter calculation module includes a multipath component elimination unit, a region division unit, and a parameter calculation unit;
[0072] The multipath component elimination unit, based on the physical laws of underwater sound propagation, determines that the propagation path length of multipath reflected signals is greater than that of direct waves, and therefore the corresponding propagation delay is greater than that of direct waves. It performs peak detection on the joint spatial-frequency distribution density matrix and eliminates peak components whose propagation delay is greater than the maximum delay of direct waves calculated based on the node and the monitoring sea area range and the average sound speed of the sea area.
[0073] The region division unit divides the spatial frequency distribution region corresponding to each scattering source according to the remaining distribution components;
[0074] The parameter calculation unit performs a weighted average of the angle sampling points for each distribution area, using the distribution density as the weight, to obtain the center angle of the scattering source. Then, it calculates the second-order central moment of the angle to obtain the angular diffusion parameters of the scattering source.
[0075] It should be noted that, according to the physical laws of underwater sound propagation, the propagation path length of multipath reflected signals is greater than that of direct waves. Therefore, the corresponding propagation delay will also be greater than that of direct waves. Peak detection is performed on the joint spatial-frequency distribution density matrix to remove peak components whose propagation delay is greater than the maximum delay of the direct wave. The maximum delay of the direct wave is calculated by dividing the maximum lateral distance of the monitored sea area by the average sound speed of the sea area. This results in the maximum propagation delay, ensuring that all multipath reflected components are removed, retaining only the direct wave component.
[0076] In the region division stage, the spatial frequency distribution region corresponding to each scattering source is divided according to the remaining distribution components.
[0077] In the parameter calculation stage, for each distribution area, the angle sampling points are weighted and averaged using the distribution density as the weight to obtain the center angle of the scattering source. Then, the second central moment of the angle is calculated to obtain the angular diffusion parameter of the scattering source. This calculation is a standard calculation method in the field of statistics and effectively obtains the distribution parameters of the scattering source.
[0078] Working principle:
[0079] The underwater acoustic signals from broadband scattering sources are collected collaboratively by distributed underwater sensor nodes. A comprehensive adaptive processing approach addresses the unique challenges of the underwater environment. Then, a low-rank reconstruction method combining spatial and frequency domains is used to achieve high-precision localization of the scattering sources. The specific workflow is as follows:
[0080] First, based on historical ocean current and topographic data of the monitored sea area, underwater sensor nodes are deployed, ensuring that the distance between nodes is less than the maximum transmission distance of the underwater acoustic communication modules used, and that the combined monitoring range of all nodes covers the target sea area. After the nodes are powered on, neighbor discovery is performed using Time-of-Flight (TOF) ranging to build a self-organizing network. Simultaneously, the Time-of-Speed (TPSN) time synchronization algorithm is used to synchronize the clocks between nodes, eliminating time discrepancies.
[0081] The distance between nodes is determined by two-way sonar ranging. Combined with the depth data of the nodes, the real-time position of the nodes is calculated, and the calibration cycle is dynamically adjusted according to the ocean current velocity to ensure the real-time position of the nodes. At the same time, the nodes collect local temperature, salinity, and depth data, and the temperature, salinity, and depth distribution of the entire sea area is obtained through linear interpolation. Then, the sound velocity profile of the entire sea area is calculated, providing parameters for subsequent direction vector correction.
[0082] After the node acquires the underwater acoustic signal from the scattering source, it performs frame segmentation and adds a Hamming window to the signal, then performs FFT transformation to the frequency domain, and then uses the Thorp attenuation model to perform amplitude compensation on the signal at different frequencies to correct the attenuation difference across the wideband. Finally, it uses inverse FFT and overlapping addition method to restore the compensated time domain signal.
[0083] The sampling matrix composed of signals from all nodes is first vectorized to obtain a standardized input vector. Then, based on the calibrated node positions and sound velocity profiles, a global cooperative reconstruction basis is constructed across nodes, where each column of the basis is the Kronecker product of the direction vector and the phase vector. Based on the reconstruction basis and the input vector, an optimization problem for low-rank matrix reconstruction is constructed. A low-rank constraint of the nuclear norm is imposed on the joint space-frequency distribution matrix, and then the optimization problem is solved using a Nesterov-accelerated proximal gradient iteration algorithm to obtain the joint space-frequency distribution density matrix.
[0084] Peak detection is performed on the distribution density matrix, multipath reflection components are eliminated based on propagation delay, the distribution area of each scattering source is divided, and finally, the center angle of the scattering source is obtained by weighted averaging with distribution density as weight, and the second-order central moment is calculated to obtain the angular diffusion parameter, thus completing the localization.
[0085] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.
Claims
1. A broadband scattering source cooperative localization system based on an underwater sensor network, characterized in that, include: Distributed underwater sensor node groups are deployed in the underwater monitoring area to collaboratively collect underwater acoustic signals from broadband scattering sources. The parameter initialization configuration module configures the system's operating parameters and completes parameter initialization before positioning. The spatial and frequency domain joint signal modeling module transforms the received signals from the distributed underwater sensor node group and establishes a joint signal model of spatial and frequency domain distribution. The collaborative reconstruction basis construction module constructs a reconstruction basis matrix adapted to multi-node collaboration based on initialization parameters and joint signal model. The sampling signal preprocessing module performs vectorization processing on the sampling matrix composed of the received signals collected by each node in row-major order to obtain a standardized input vector. The low-rank constraint optimization solution module constructs a low-rank matrix reconstruction optimization problem based on the reconstructed basis and input vector, and applies a nuclear norm low-rank constraint to the joint space-frequency distribution matrix to complete the problem solution. The scattering source parameter calculation module calculates the location parameters of the broadband scattering source based on the obtained spatial-frequency joint distribution density matrix. The distributed underwater sensor node group includes a node deployment unit, a self-organizing networking unit, a dynamic position calibration unit, and an underwater acoustic clock synchronization unit; The node deployment unit reads the pre-stored historical ocean current distribution data and topographic data of the monitoring area, and deploys the underwater sensor nodes in the target sea area. The distance between the nodes is set to be less than the maximum transmission distance of the underwater acoustic communication module used. The union of the monitoring ranges of all nodes covers the entire target sea area. The self-organizing networking unit broadcasts a neighbor discovery request after the node is powered on. After receiving the request, the neighbor node obtains the distance to the neighbor node through TOF ranging, replies with a response, builds a communication route, and forms a self-organizing network. Each underwater sensor node integrates an underwater acoustic transducer, a signal acquisition unit, and a short-range underwater acoustic communication unit. The communication unit adjusts the transmission power according to the distance to the neighbor node measured during the neighbor discovery process. The greater the distance, the greater the transmission power. The dynamic position calibration unit acquires bidirectional sonar ranging data obtained through two-way sonar interaction between nodes, reads node depth data collected by the depth sensors on each node, calculates the real-time position of the node based on the ranging data and depth data, updates the node's position parameters to the calculated real-time position, and adjusts the calibration cycle according to the real-time ocean current velocity of the monitored sea area; the higher the current velocity, the shorter the calibration cycle. The underwater acoustic clock synchronization unit works as follows: Node A sends a data packet carrying a local timestamp to Node B. Node B records its local time when it receives the data packet. Node B replies to Node A with a data packet carrying the received time and its local timestamp. Node A records its local time when it receives the data packet. The clock deviation between the nodes is calculated based on the four timestamps to complete the clock deviation calibration between the distributed nodes. The parameter initialization configuration module includes an array parameter configuration unit, a search range configuration unit, a discrete precision configuration unit, and a sound velocity profile adaptation unit. The array parameter configuration unit reads the position parameters of each underwater sensor node after dynamic calibration and completes the parameter configuration of the distributed node array. The search range configuration unit configures the target angle search range according to the lateral range of the monitored sea area, and configures the system's operating wideband range according to the frequency range of the radiation signals from the pre-stored scattering sources. The discrete precision configuration unit configures the step size parameter of spatial frequency domain discretization according to the preset maximum positioning error. The smaller the preset maximum positioning error, the smaller the corresponding step size. The preset maximum positioning error ranges from 0.01 degrees to 1 degree. The sound velocity profile adaptation unit loads real-time temperature, salinity, and depth data collected by the temperature, salinity, and depth sensors built into each underwater sensor node. It obtains the temperature, salinity, and depth distribution of the entire monitored sea area through linear interpolation, calculates the sound velocity profile data of the entire sea area based on the empirical formula for sound velocity, and corrects the propagation delay parameter of the direction vector based on the sound velocity profile, providing sound propagation parameters for the calculation of the direction vector. The global collaborative reconstruction basis uses discrete sampling points as columns. Each column corresponds to a combination of angle sampling points and frequency sampling points. The elements of each column are the Kronecker product with the joint direction vector corrected for the speed of sound first and the time sampling phase vector second. The time sampling phase vector is the phase sequence corresponding to the time sampling points after clock synchronization.
2. The broadband scattering source cooperative localization system based on an underwater sensor network according to claim 1, characterized in that, The spatial-frequency domain joint signal modeling module includes a frequency domain transformation unit, a wideband attenuation compensation unit, and a time domain reconstruction unit. The frequency domain transformation unit performs frame processing on the received time domain signal of each underwater sensor node, sets the frame length according to the signal length, performs windowing processing on the signal using a Hamming window, performs amplitude compensation on the windowed signal, sets the FFT length to the same as the frame length, and performs fast Fourier transform to convert the time domain signal into a frequency domain signal. The wideband attenuation compensation unit calculates the attenuation coefficient corresponding to different frequencies based on the Thorp attenuation model, and multiplies the amplitude of the frequency domain signal by the reciprocal of the corresponding attenuation coefficient to complete the amplitude compensation. The time-domain reconstruction unit performs an inverse Fourier transform on the compensated frequency-domain signal and restores the framed signal to a complete time-domain signal through the overlapping addition method.
3. A broadband scattering source cooperative localization system based on an underwater sensor network according to claim 2, characterized in that, After processing by the frequency domain transformation unit, the frequency domain signal satisfies the following: for each frequency point, the frequency domain signal is equal to the sum of the spatial integral results of all scattering sources and the environmental noise. The direction vector in the integral term is the multi-node joint direction vector after sound speed profile correction, the integration variable is the target angle, and the integrand is the product of the direction vector and the spatial frequency distribution density of the corresponding scattering source.
4. A broadband scattering source cooperative localization system based on an underwater sensor network according to claim 3, characterized in that, After processing by the time-domain reconstruction unit, the time-domain signal satisfies the following condition: for each time point, the time-domain signal is equal to the sum of the spatial-frequency double integral results of all scattering sources and the time-domain noise. The integral term includes the joint direction vector corrected for sound speed, the Thorp attenuation compensation coefficient, the spatial-frequency distribution density, and the phase term. The integral variables are the target angle and frequency.
5. A broadband scattering source cooperative localization system based on an underwater sensor network according to claim 4, characterized in that, The collaborative reconstruction basis construction module includes dimension discretization units, local manifold computation units, and cross-node fusion units; The dimension discretization unit reads the initialized discretization step size parameters and performs grid discretization on the angle dimension and frequency dimension to obtain angle sampling points and frequency sampling points; The local manifold computation unit calculates the local array manifold vector of each underwater sensor node for each discrete sampling point based on the calibrated node position and sound velocity profile. The cross-node fusion unit splices together the local manifold vectors of all nodes to construct a global collaborative reconstruction basis across nodes.
6. A broadband scattering source cooperative localization system based on an underwater sensor network according to claim 1, characterized in that, The low-rank constraint optimization solution module includes a problem transformation unit, an accelerated iterative solution unit, and a convergence verification unit; The problem transformation unit constructs an optimization problem by combining the vectorized preprocessed input vector with the collaborative reconstruction basis. The original rank minimization problem is transformed into a least squares optimization problem with a rank upper limit constraint. The rank upper limit is set according to the maximum number of scattering sources preset by the monitoring task. The rank of the space-frequency joint distribution matrix and the number of scattering sources have a 1:1 correspondence. The accelerated iterative solution unit employs a Nesterov-accelerated proximal gradient iteration algorithm for the nuclear norm to iteratively solve the optimization problem; The convergence verification unit calculates the error for each iteration. When the iteration error is less than the convergence threshold, the iteration is terminated, and the joint spatial frequency distribution density matrix is output. The smaller the preset maximum positioning error, the smaller the corresponding convergence threshold.
7. A broadband scattering source cooperative localization system based on an underwater sensor network according to claim 6, characterized in that, The scattering source parameter calculation module includes a multipath component removal unit, a region partitioning unit, and a parameter calculation unit; The multipath component elimination unit, based on the physical laws of underwater sound propagation, determines that the propagation path length of multipath reflected signals is greater than that of direct waves, and therefore the corresponding propagation delay is greater than that of direct waves. It performs peak detection on the joint spatial-frequency distribution density matrix and eliminates peak components whose propagation delay is greater than the maximum delay of direct waves calculated based on the node and the monitoring sea area range and the average sound speed of the sea area. The region division unit divides the spatial frequency distribution region corresponding to each scattering source according to the remaining distribution components; The parameter calculation unit performs a weighted average of the angle sampling points for each distribution area, using the distribution density as the weight, to obtain the center angle of the scattering source. Then, it calculates the second-order central moment of the angle to obtain the angular diffusion parameters of the scattering source.