An underwater obstacle avoidance system for unmanned boats based on distributed wide-beam sonar

Abstract

The present invention discloses an underwater obstacle avoidance system for an unmanned boat based on distributed wide-beam sonar. The system collects original environmental information through a wide-beam sonar group and uses adaptive filtering to perform dynamic noise reduction processing to obtain environmental detection information. The navigation compensation module obtains the motion state of the unmanned boat and performs motion error compensation on the environmental detection information to obtain environmental correction information. Based on the environmental correction information, an ellipsoidal space safety area is constructed and the safe navigation boundary is calculated in real time to obtain safe navigation information. The obstacle avoidance decision module generates a multi-level obstacle avoidance strategy based on the real-time navigation status and safe navigation information. The execution control module converts the strategy into an obstacle avoidance instruction and dynamically adjusts the output frequency to adapt to different navigation conditions. The system effectively solves the problems of insufficient detection accuracy and delayed response of traditional obstacle avoidance systems in complex underwater environments, realizes all-round and high-precision control of underwater obstacle avoidance of unmanned boats, and significantly improves the real-time and environmental adaptability of the obstacle avoidance system.

Description

Technical Field

[0001] The present invention relates to the technical field of unmanned boats, and in particular to an underwater obstacle avoidance system for an unmanned boat based on distributed wide-beam sonar. Background Art

[0002] With the increasing use of unmanned aerial vehicles (UAVs) in underwater operations, the challenges they face in complex underwater environments are becoming increasingly severe. Autonomous obstacle avoidance systems are a core technology for ensuring navigation safety, and sonar detection has become a mainstream solution due to its all-weather operation. Traditional systems often use centralized, co-located sonar transmitters and receivers, achieving environmental awareness through mechanical rotation or phased array technology. However, in actual operations, due to the complex underwater environment and the physical limitations of the equipment, current technical bottlenecks include: In dynamic sea conditions, the violent hull swing causes sonar beam misalignment, which, combined with shallow water reverberation (e.g., a 70% reduction in effective detection range at a depth of 5 meters), distorts obstacle feature extraction. The inherent electromagnetic coupling of the co-located transmitter and receiver creates a near-field blind spot (typically 0.35-1 meter), which, along with the response delay of the mechanical scanning mechanism (≥0.5 seconds), limits rapid obstacle avoidance. Frequency band conflicts when multiple sonars operate in concert (e.g., 137kHz interference reduces detection range by 40%) and the inherent latency of centralized processing (full link cycle exceeding 500ms) make it difficult to meet the real-time obstacle avoidance requirements during high-speed navigation. These factors collectively result in an obstacle misjudgment rate exceeding 12% for existing systems in complex waters, severely limiting the reliability of unmanned vehicles in typical scenarios such as island reefs and shallow shoals. Summary of the Invention

[0003] In view of this, the purpose of the present invention is to propose an unmanned boat underwater obstacle avoidance system based on distributed wide-beam sonar to solve the problems of insufficient sonar detection accuracy and real-time obstacle avoidance capability in complex underwater environments.

[0004] In order to achieve the above technical objectives, the technical solution adopted by the present invention is: an underwater obstacle avoidance system for unmanned boats based on distributed wide-beam sonar, including a sonar detection module, a navigation compensation module, a safety domain construction module, an obstacle avoidance decision module and an execution control module. The sonar detection module includes a wide-beam sonar group, each wide-beam sonar group includes a wide-beam transmitter and a wide-beam receiver. The sonar detection module is used to collect original environmental information and use adaptive filtering to dynamically reduce the noise of the original environmental information to obtain environmental detection information; the navigation compensation module is used to obtain the motion state of the unmanned boat and detect the environmental information. The information is used to compensate for motion errors and obtain environmental correction information; the safety domain construction module is used to construct the ellipsoid space safety area of ​​the current unmanned boat according to the environmental correction information, and calculate the safe navigation boundary corresponding to the ellipsoid space safety area in real time to obtain safe navigation information; the obstacle avoidance decision module is used to obtain the real-time navigation status and safe navigation information of the current unmanned boat, and generate a multi-level obstacle avoidance strategy with priority; the execution control module is used to convert the multi-level obstacle avoidance strategy into obstacle avoidance instructions, which include heading control instructions and speed control instructions, and dynamically adjust the output frequency of the obstacle avoidance instructions to adapt to different navigation conditions.

[0005] In some embodiments, the sonar detection module is used to collect raw environmental information and dynamically reduce noise on the raw environmental information using adaptive filtering to obtain environmental detection information, including:

[0006] Control the wide beam transmitter to transmit sound wave signals. The wide beam transmitter adopts linear frequency modulation. The transmission power of the wide beam transmitter is generated according to the reverberation intensity monitored in real time.

[0007] The underwater echo signal is received by a wide-beam receiver, and the underwater echo signal is amplified, band-pass filtered, and analog-to-digital converted to obtain the original environmental information;

[0008] Adaptive filtering is performed on the original environmental information, including:

[0009] Performing primary filtering on the original environmental information using a normalized least mean square algorithm to obtain first filtered information;

[0010] Performing secondary filtering on the first filtered information using a convolutional neural network model to obtain second filtered information;

[0011] A pulse compression algorithm is used to estimate the time delay of the second filter information, and the earliest echo arrival time is corrected to obtain the third filter information;

[0012] Calculating the obstacle distance based on the sound speed and the third filter information, and generating environment detection information including the obstacle position, obstacle distance, and obstacle strength;

[0013] Also, in the event of communication anomalies, the environmental detection information is temporarily stored in the local cache, and the data collection timestamp and signal quality indicators are recorded.

[0014] In some embodiments, the convolutional neural network model is trained by the following steps:

[0015] Construct an initial network model and input sample data for training. The sample data includes sample echo signals. The initial network model includes a feature extraction layer and a feature classification layer. The feature extraction layer is used to perform multi-scale feature extraction on the sample echo signals to obtain sample target echoes and sample noise components. The feature classification layer is used to distinguish between sample target echoes and sample noise components.

[0016] The initial network model is trained repeatedly until the classification accuracy reaches the preset threshold to obtain a convolutional neural network model.

[0017] In some embodiments, the navigation compensation module is used to obtain the motion state of the unmanned vehicle and perform motion error compensation on the environmental detection information to obtain environmental correction information, including:

[0018] The real-time attitude data of the unmanned boat is collected through the MEMS inertial measurement unit, including roll angle, pitch angle and heading angle;

[0019] The position and speed information of the unmanned boat is obtained through a dual-frequency GNSS receiver, and a tightly coupled navigation solution model is constructed by combining the position information, speed information and real-time attitude data;

[0020] Based on the improved robust UKF algorithm, data fusion is performed on the tightly coupled navigation solution model to calculate the motion compensation parameters of the unmanned boat in the world coordinate system. The motion compensation parameters include position offset and attitude correction.

[0021] The coordinate transformation compensation of the environmental detection information is performed according to the motion compensation parameters, including:

[0022] Establish the conversion relationship between the hull coordinate system and the world coordinate system;

[0023] Convert the obstacle distance in the environment detection information from the ship coordinate system to the world coordinate system;

[0024] Compensate for sonar measurement errors caused by ship motion;

[0025] Outputs environmental correction information containing the precise position of obstacles in the world coordinate system;

[0026] Also, when the GNSS signal is lost, it switches to pure inertial navigation mode and records the navigation status flag.

[0027] In some embodiments, the safety domain construction module is used to construct the ellipsoidal space safety area of ​​the current unmanned vehicle based on the environmental correction information, including:

[0028] Obtain the obstacle positions in the environmental correction information one by one, and extract the azimuth and distance information of the obstacles relative to the unmanned boat;

[0029] The first actual position of the wide beam transmitter and the second actual position of the wide beam receiver at the current moment are used as dual focal points, and the ellipsoid parameters are calculated based on the sound wave propagation time difference, including:

[0030] The shortest propagation distance corresponding to the original environmental information of the earliest echo is taken as the major axis of the ellipsoid;

[0031] Determine the focal length of the ellipsoid according to the distance between the transmitting and receiving transducers, and calculate the minor axis of the ellipsoid according to the real-time draft depth of the unmanned boat;

[0032] Obtain the ellipsoid parameters according to the ellipsoid major axis, ellipsoid minor axis and ellipsoid focal length;

[0033] A dynamic ellipsoid safety model is constructed based on the ellipsoid parameters in the world coordinate system, including:

[0034] Establish the equation of the ellipsoid with the transmitting and receiving transducers as the dual foci;

[0035] Update the position in ellipsoid space according to real-time navigation data;

[0036] Visualize the safe area in ellipsoid space.

[0037] In some embodiments, the safe navigation boundary corresponding to the safe area in the ellipsoid space is calculated in real time to obtain safe navigation information, including:

[0038] Calculate the spatial relationship between the ellipsoidal space safety area and the unmanned boat's hull outline to generate the initial safe navigation boundary;

[0039] Outputting an initial safety area dataset of the ellipsoid space safety area according to the initial safety navigation boundary, the initial safety area dataset including an initial boundary coordinate set and an initial nearest obstacle warning level;

[0040] Based on the initial boundary coordinate set, a three-dimensional space grid map is established, where the three-dimensional space grid map includes a plurality of grid units;

[0041] Mark grid cells one by one as safe area, warning area or dangerous area according to the initial nearest obstacle warning level;

[0042] Obtain the dynamic parameters of the unmanned boat and calculate the navigable area based on the dynamic parameters. The dynamic parameters include the current speed, steering angle, roll angle, and pitch angle of the unmanned boat, including:

[0043] Generate the minimum turning radius according to the current speed and steering angle of the unmanned boat;

[0044] Generate the influence coefficient of the current minimum turning radius on the actual operable space based on the roll angle and pitch angle;

[0045] Generate the navigable area of ​​the three-dimensional grid map based on the minimum turning radius, influence coefficient and preset safety margin threshold;

[0046] Perform multi-level boundary analysis within the navigable domain, including:

[0047] Calculate the minimum safe distance from the ellipsoid surface to the nearest obstacle to obtain the primary boundary;

[0048] The track envelope within a preset short time interval is predicted based on the primary boundary and the current speed to generate the secondary boundary.

[0049] Generate dynamic safety corridors based on primary boundaries and secondary boundaries to obtain safe navigation boundaries;

[0050] Also, when it is detected that the width of the dynamic safety corridor is less than a preset multiple of the hull width, the speed reduction mode is automatically triggered and the safe navigation boundary is replanned;

[0051] Based on the safe navigation boundary, safe navigation information is generated, which includes the optimal safe navigation direction recommendation, the maximum allowable speed in each direction, the emergency braking distance parameters, and the multi-level warning status identification.

[0052] In some embodiments, the obstacle avoidance decision module is used to obtain the real-time navigation status and safe navigation information of the current unmanned vehicle and generate a multi-level obstacle avoidance strategy with priority, including:

[0053] Construct an obstacle avoidance decision tree based on risk assessment. The obstacle avoidance decision tree includes a first-level response decision layer, a second-level response decision layer, and a third-level response decision layer. The first-level response decision layer is configured to immediately trigger an emergency braking command when the collision time is less than 3 seconds. The second-level response decision layer is configured to generate a course correction strategy when the width of the dynamic safety corridor is less than 2 times the ship width. The third-level response decision layer is configured to select an obstacle avoidance route based on the principle of optimal energy consumption when there are multiple feasible paths.

[0054] The real-time navigation status and safe navigation information of the current unmanned boat are input into the obstacle avoidance decision tree to obtain the output results, and the output results are dynamically optimized, including:

[0055] Real-time calculation of the feasibility index of each obstacle avoidance solution in the output results. The feasibility index includes the path smoothness score, energy consumption cost score, and task delay cost score.

[0056] The feasibility index is weighted and integrated using fuzzy logic algorithm to obtain the comprehensive score of each obstacle avoidance solution;

[0057] The priorities of multiple obstacle avoidance schemes are generated according to the comprehensive scoring scores, and a multi-level obstacle avoidance strategy package is generated.

[0058] In some embodiments, the execution control module is used to convert the multi-level obstacle avoidance strategy into obstacle avoidance instructions, where the obstacle avoidance instructions include heading control instructions and speed control instructions, including:

[0059] Convert the path planning parameters of each obstacle avoidance path in the multi-level obstacle avoidance strategy into heading control instructions, including:

[0060] Extract the heading angle sequence of the obstacle avoidance path;

[0061] A trajectory smoothing algorithm is used to generate continuous steering commands based on the heading angle sequence;

[0062] Calculate the actual rudder angle control value according to the rudder effect characteristics and obtain the heading control instruction;

[0063] Convert the speed planning parameters of each obstacle avoidance path in the multi-level obstacle avoidance strategy into speed control instructions, including:

[0064] Analyze the recommended speed under the safety distance constraint and generate the acceleration curve that meets the propulsion characteristics;

[0065] Calculate the final thrust control parameters according to the acceleration curve and obtain the speed control command;

[0066] Obstacle avoidance instructions are generated based on the heading control instructions and speed control instructions, and are output through the multi-mode output interface. The multi-mode output interface is configured as follows:

[0067] In normal cruise mode, the periodic command output mode is adopted;

[0068] Enable event-triggered command output mode in emergency obstacle avoidance mode;

[0069] Switch to degraded instruction output mode when the system is abnormal;

[0070] Also, safety verification is performed on obstacle avoidance instructions before output, including:

[0071] Verify the valid range of the command parameters of the obstacle avoidance command;

[0072] Limit the command change rate of obstacle avoidance commands;

[0073] Check the coordination between multiple instructions of obstacle avoidance instructions.

[0074] In some embodiments, the output frequency of obstacle avoidance instructions is dynamically adjusted to adapt to different navigation conditions, including:

[0075] Determine the reference output frequency based on the current navigation conditions, including:

[0076] Obtain the real-time navigation status parameters of the unmanned boat;

[0077] Identify the current operating condition type based on real-time navigation status parameters;

[0078] Query the working condition-frequency mapping table to determine the reference output frequency;

[0079] When the speed of the real-time navigation status parameter increases, the reference output frequency is gradually increased, and the frequency adjustment log information is recorded;

[0080] Perform smooth transition processing during reference output frequency adjustment, including:

[0081] Set the frequency switching transition interval and adopt a gradual adjustment algorithm to maintain the continuity of the obstacle avoidance command output process.

[0082] In some embodiments, the unmanned boat underwater obstacle avoidance system based on distributed wide-beam sonar also includes a reverberation suppression module and a zero-range blind zone compensation module. The reverberation suppression module is used to establish a seabed-water surface multipath reflection channel model, use an adaptive notch filter to suppress reverberation in a specific frequency band, eliminate false echoes, and adjust the signal pulse width of the transmitted signal to optimize the distance resolution; the zero-range blind zone compensation module is used to establish an error compensation model based on hull vibration, integrate inertial data to predict the position of proximal obstacles, and use a time-gated algorithm to isolate self-interference signals.

[0083] By adopting the above-mentioned technical solution, the present invention has the following beneficial effects compared with the prior art: the present invention provides an underwater obstacle avoidance system for an unmanned boat based on distributed wide-beam sonar, which collects original environmental information through a wide-beam sonar group and uses adaptive filtering to perform dynamic noise reduction processing to obtain environmental detection information. The navigation compensation module obtains the motion state of the unmanned boat and performs motion error compensation on the environmental detection information to obtain environmental correction information. Based on the environmental correction information, an ellipsoidal space safety area is constructed and the safe navigation boundary is calculated in real time to obtain safe navigation information. The obstacle avoidance decision module generates a multi-level obstacle avoidance strategy based on the real-time navigation status and safe navigation information, and the execution control module converts the strategy into an obstacle avoidance instruction and dynamically adjusts the output frequency to adapt to different navigation conditions. The system effectively solves the problems of insufficient detection accuracy and delayed response of traditional obstacle avoidance systems in complex underwater environments, realizes all-round and high-precision control of underwater obstacle avoidance of unmanned boats, and significantly improves the real-time and environmental adaptability of the obstacle avoidance system. BRIEF DESCRIPTION OF THE DRAWINGS

[0084] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0085] Figure 1 Schematic diagram of the layout of the distributed wide-beam sonar on the unmanned boat according to the specific implementation method;

[0086] Figure 2 is a schematic diagram of the ellipsoidal space safety area described in the specific implementation method;

[0087] Figure 3 Schematic diagram of the relationship between the wide-beam sonar beam angle and the unmanned boat described in the specific implementation method;

[0088] Figure 4 Schematic diagram of the dynamic activity space and ellipsoidal space safety area of ​​the unmanned boat described in the specific implementation method;

[0089] Figure 5 It is a schematic diagram of distance calculation for the dispersed deployment of wide beam transmitters described in a specific implementation method.

[0090] The reference numerals are as follows:

[0091] 1. Unmanned boat;

[0092] 2. Wide-beam sonar set;

[0093] 21. Wide beam transmitter;

[0094] 22. Wide beam receiver;

[0095] 3. Ellipsoid space safety area;

[0096] 4. Dynamic activity space;

[0097] a. Wide beam transmitter beam angle;

[0098] b. Wide beam receiver beam angle. DETAILED DESCRIPTION

[0099] The present invention will be described in further detail below with reference to the accompanying drawings and examples. It is particularly noted that the following examples are intended only to illustrate the present invention and are not intended to limit the scope of the present invention. Similarly, the following examples are only some embodiments of the present invention and are not intended to be exhaustive. All other embodiments obtained by those of ordinary skill in the art without creative effort are intended to fall within the scope of protection of the present invention.

[0100] See also Figures 1 to 5This embodiment provides an underwater obstacle avoidance system for an unmanned boat based on distributed wide-beam sonar, including a sonar detection module, a navigation compensation module, a safety domain construction module, an obstacle avoidance decision module, and an execution control module. The sonar detection module includes a wide-beam sonar group 2, each wide-beam sonar group 2 includes a wide-beam transmitter 21 and a wide-beam receiver 22. The sonar detection module is used to collect original environmental information and use adaptive filtering to dynamically reduce noise on the original environmental information to obtain environmental detection information; the navigation compensation module is used to obtain the motion state of the unmanned boat 1 and perform motion error correction on the environmental detection information. The safety domain construction module is used to construct the ellipsoid space safety area 3 of the current unmanned boat 1 according to the environmental correction information, and calculate the safe navigation boundary corresponding to the ellipsoid space safety area 3 in real time to obtain safe navigation information; the obstacle avoidance decision module is used to obtain the real-time navigation status and safe navigation information of the current unmanned boat 1, and generate a multi-level obstacle avoidance strategy with priority; the execution control module is used to convert the multi-level obstacle avoidance strategy into obstacle avoidance instructions, which include heading control instructions and speed control instructions, and dynamically adjust the output frequency of the obstacle avoidance instructions to adapt to different navigation conditions.

[0101] In this embodiment, the wide beam sonar group 2 adopts a sonar detection and emission separation design, which is physically isolated and spaced apart. The wide beam transmitter 21 (piezoelectric / magnetostrictive / electromagnetic) and the wide beam receiver 22 are composed of a wide beam transmitter 21 (piezoelectric / magnetostrictive / electromagnetic) and a wide beam receiver 22. Preferably, the wide beam transmitter beam angle a and the wide beam receiver beam angle b are both 150 degrees and meet the condition of not directly irradiating the bottom of the ship. The wide beam transmitter 21 and the wide beam receiver 22 are arranged in a The layout rule is installed at the bottom of the ship's side, and the independent transmitting circuit is connected to the signal processing circuit including the amplification / filtering circuit through a waterproof cable. is the maximum detection distance of the ultrasonic radar, is the maximum safe area distance that needs to be measured, that is, the distance from the target to the unmanned boat 1. The value is 0.4-0.6.

[0102] For details, please refer to Figure 5 In the figure, C and D are the positions of the wide-beam sonar group 2, E is the target, and F is the intersection of the shortest distance between the target E and the unmanned boat. Preferably, a wide-beam receiver 22 and a wide-beam transmitter 21 are arranged at the bow and stern of the bottom of the ship on both sides of the unmanned boat 1 to achieve the separation of transmission and reception and meet the requirements of the spacing between the transceiver and the receiver: when the maximum safe area distance is EF and the maximum detection distance of the ultrasonic radar is R, the square of the spacing CD between the transceiver and the receiver is less than When a single wide-beam transmitter 21 is composed of a single wide-beam receiver 22, they must be separated. When multiple wide-beam receivers 22 or wide-beam transmitters 21 are used, the wide-beam receivers 22 and wide-beam receivers 22, and the wide-beam transmitters 21 and wide-beam transmitters 21 are separated. The wide-beam transmitter 21 and wide-beam receiver 22 can be combined. The edge transmission line (the line where the energy decays to 1 / 2 of the maximum value) of the wide-beam transmitter beam angle a and the wide-beam receiver beam angle b should not be directly transmitted to the bottom of the ship to generate reflected echoes and avoid detection errors. The beam angle should be as large as possible, and the coverage solid angle is determined by the bottom structure of the ship.

[0103] When collecting original environmental information, an asynchronous co-frequency or synchronous heterodyne modulation signal transmission strategy is adopted. After the receiver detects the reflected sound waves, dynamic noise reduction is performed through adaptive filtering based on the NLMS algorithm. The filter characteristics are automatically adjusted according to the real-time reverberation intensity to retain effective obstacle echoes, and ultimately output environmental detection information with an improved signal-to-noise ratio.

[0104] The navigation compensation module uses a tightly coupled system consisting of a MEMS-IMU and a dual-frequency GNSS receiver to acquire the UAV's motion state (including roll angle, heading angle, and acceleration). This data is then fused using an improved robust UKF algorithm based on SVD. Motion error compensation is achieved by first converting the sonar coordinate system to the Earth coordinate system and then applying a kinematic model to eliminate beam pointing deviations caused by vessel motion, keeping the ranging error of the environmental correction information within 0.3 meters.

[0105] Ellipsoid safety zone 3 is constructed in the world coordinate system, centered on the real-time positions of the transmitter and receiver. The shortest sound wave propagation distance is used as the major axis of the ellipsoid, and the radius of the minor axis is adaptively adjusted based on the current speed. The safe navigation boundary is determined by calculating the minimum distance between the ellipsoid surface and the obstacle point cloud. When the boundary threshold is exceeded, the obstacle avoidance strategy is triggered.

[0106] During the generation of multi-level obstacle avoidance strategies, priorities are assigned based on three dimensions: obstacle distance, relative speed, and threat level. Heading control commands are quantified through servo deflection angles, while speed control commands are converted into propulsion motor speed percentages. The output frequencies of both are dynamically adjusted based on water depth (10Hz in shallow water, 5Hz in open water).

[0107] The connection between the execution control module and the power system utilizes hardware-triggered synchronization, ensuring time synchronization accuracy of less than 0.1 millisecond and maintaining an overall response latency of less than 200 milliseconds at a speed of 15 knots. All modules are collaboratively controlled by a processing unit based on an industrial-grade MCU, ultimately achieving real-time matching between the unmanned vehicle's (1) dynamic activity space (4) and the ellipsoidal safety zone (3).

[0108] This embodiment effectively improves the reliability of environmental detection information in complex underwater environments through the non-directional wide-angle detection characteristics of the distributed wide-beam sonar group 2, combined with the dynamic noise reduction processing of adaptive filtering. The navigation compensation module performs motion error compensation based on the motion state of the unmanned boat 1 to ensure the spatial accuracy of the environmental correction information. The safety domain construction module realizes the precise definition of the safe navigation boundary through the dynamic calculation of the ellipsoid space safety area 3. The multi-level obstacle avoidance strategy generated by the obstacle avoidance decision module is coordinated with the heading control instructions and speed control instructions of the execution control module, so that the system can adaptively adjust the output frequency according to different navigation conditions, and ultimately achieve all-round and high-precision control of the underwater obstacle avoidance of the unmanned boat 1. The modules work together to significantly improve the real-time and environmental adaptability of the obstacle avoidance system while keeping the hardware structure simple.

[0109] In some embodiments, the sonar detection module is used to collect raw environmental information and dynamically reduce noise on the raw environmental information using adaptive filtering to obtain environmental detection information, including:

[0110] Control the wide beam transmitter to transmit sound wave signals. The wide beam transmitter adopts linear frequency modulation. The transmission power of the wide beam transmitter is generated according to the reverberation intensity monitored in real time.

[0111] The underwater echo signal is received by a wide-beam receiver, and the underwater echo signal is amplified, band-pass filtered, and analog-to-digital converted to obtain the original environmental information;

[0112] Adaptive filtering is performed on the original environmental information, including:

[0113] Performing primary filtering on the original environmental information using a normalized least mean square algorithm to obtain first filtered information;

[0114] Performing secondary filtering on the first filtered information using a convolutional neural network model to obtain second filtered information;

[0115] A pulse compression algorithm is used to estimate the time delay of the second filter information, and the earliest echo arrival time is corrected to obtain the third filter information;

[0116] Calculating the obstacle distance based on the sound speed and the third filter information, and generating environment detection information including the obstacle position, obstacle distance, and obstacle strength;

[0117] Also, in the event of communication anomalies, the environmental detection information is temporarily stored in the local cache, and the data collection timestamp and signal quality indicators are recorded.

[0118] In this embodiment, the wide-beam transmitter 21 adopts linear frequency modulation, and its transmission power is dynamically adjusted based on real-time reverberation intensity feedback. When the reverberation intensity exceeds a preset threshold, the transmission power is automatically reduced to avoid signal saturation. Otherwise, the power is increased to ensure the detection distance.

[0119] Preferably, the wide beam receiver 22 achieves echo optimization through a three-stage signal processing chain: the signal amplification process uses a programmable gain amplifier, whose gain coefficient is inversely proportional to the received signal strength; the cutoff frequency of the bandpass filtering process is set according to the frequency band characteristics of the transmitted signal; the analog-to-digital conversion process uses oversampling technology to improve the signal-to-noise ratio.

[0120] Adaptive filtering of raw environmental information utilizes a hierarchical, progressive architecture. A normalized least mean square (NLMS) algorithm suppresses steady-state noise with a variable step-size parameter, adaptively matching its convergence speed to the signal's dynamic range. A convolutional neural network model employs a three-layer convolutional architecture, extracting and filtering features from non-stationary interference using an offline-trained weight matrix. A pulse compression algorithm improves delay estimation accuracy through matched filtering, with a compression ratio proportional to the time-width-bandwidth product of the transmitted signal. The resulting third-filtered information is used to calculate obstacle distances using a speed-of-sound-delay conversion model, where the speed of sound is adjusted in real time based on water temperature and salinity.

[0121] Environmental detection information includes obstacle location, distance, and strength. Preferably, obstacle location is resolved using a beamforming algorithm, and obstacle strength is logarithmically compressed. A dual-buffer design is used to temporarily store data during communication anomalies. Data acquisition timestamps are generated synchronously with the hardware clock. Signal quality indicators include dynamic range and harmonic distortion, providing reliability criteria for subsequent data fusion.

[0122] This embodiment effectively adapts to diverse reverberation environments through dynamic power regulation of the wide-beam transmitter 21 and a three-stage signal processing chain of the wide-beam receiver 22. Adaptive filtering, including a normalized least mean square algorithm, a convolutional neural network model, and a pulse compression algorithm, achieves hierarchical suppression of steady-state and non-steady-state noise. A speed-of-sound-delay conversion model and beamforming algorithm accurately analyze obstacle information. A double buffer design and signal quality indicator recording during communication anomalies ensure data reliability. This embodiment significantly improves the accuracy and robustness of underwater environment detection.

[0123] In some embodiments, the convolutional neural network model is trained by the following steps:

[0124] Construct an initial network model and input sample data for training. The sample data includes sample echo signals. The initial network model includes a feature extraction layer and a feature classification layer. The feature extraction layer is used to perform multi-scale feature extraction on the sample echo signals to obtain sample target echoes and sample noise components. The feature classification layer is used to distinguish between sample target echoes and sample noise components.

[0125] The initial network model is trained repeatedly until the classification accuracy reaches the preset threshold to obtain a convolutional neural network model.

[0126] In this embodiment, the sample echo signal is the original acoustic signal collected by the wide-beam receiver 22, which includes target reflections and various types of interference. It is acquired through a combination of underwater measurements and simulation modeling. The feature extraction layer implements multi-scale analysis using convolution kernels of different sizes. Large-scale convolution kernels are used to extract the macroscopic waveform features of the sample target echo, while small-scale convolution kernels are used to capture the local perturbation patterns of the sample noise component. The feature classification layer uses a fully connected network structure and outputs a binary discrimination result between the sample target echo and the sample noise component using a sigmoid activation function.

[0127] Preferably, the preset threshold is set according to the misjudgment tolerance of the actual application scenario, and its determination process is evaluated by the recall-precision curve of the validation set. When the balance point of the two reaches the engineering requirements, the training is terminated. Preferably, the small batch gradient descent method is used for repeated training, and the classification accuracy of the validation set is calculated after each iteration. The model optimization is completed when the accuracy fluctuation of three consecutive training cycles is less than the convergence criterion. The parameters of the convolutional neural network model finally obtained are frozen and embedded in the signal processing circuit, and the noise component is identified and filtered in real time in the secondary filtering stage.

[0128] This embodiment uses a multi-scale feature extraction layer to perform hierarchical analysis of sample echo signals, combined with a feature classification layer to accurately distinguish target echoes from noise components. This enables the trained convolutional neural network model to possess adaptive noise suppression capabilities. An iterative training mechanism controlled by preset thresholds ensures model convergence, ultimately achieving improved classification accuracy and effectively ensuring the reliable recognition of underwater target echo signals.

[0129] In some embodiments, the navigation compensation module is used to obtain the motion state of the unmanned vehicle and perform motion error compensation on the environmental detection information to obtain environmental correction information, including:

[0130] The real-time attitude data of the unmanned boat is collected through the MEMS inertial measurement unit, including roll angle, pitch angle and heading angle;

[0131] The position and speed information of the unmanned boat is obtained through a dual-frequency GNSS receiver, and a tightly coupled navigation solution model is constructed by combining the position information, speed information and real-time attitude data;

[0132] Based on the improved robust UKF algorithm, data fusion is performed on the tightly coupled navigation solution model to calculate the motion compensation parameters of the unmanned boat in the world coordinate system. The motion compensation parameters include position offset and attitude correction.

[0133] The coordinate transformation compensation of the environmental detection information is performed according to the motion compensation parameters, including:

[0134] Establish the conversion relationship between the hull coordinate system and the world coordinate system;

[0135] Convert the obstacle distance in the environment detection information from the ship coordinate system to the world coordinate system;

[0136] Compensate for sonar measurement errors caused by ship motion;

[0137] Outputs environmental correction information containing the precise position of obstacles in the world coordinate system;

[0138] Also, when the GNSS signal is lost, it switches to pure inertial navigation mode and records the navigation status flag.

[0139] In this embodiment, the MEMS inertial measurement unit (IMU) preferably utilizes an industrial-grade microelectromechanical system (e.g., the SBG Apogee series). It measures the real-time motion of the unmanned vehicle 1 through a combination of a three-axis fiber optic gyroscope and a quartz accelerometer. The roll, pitch, and heading angles represent the changes in the hull's lateral tilt, longitudinal tilt, and azimuth orientation, respectively. These measurement data are output after temperature compensation and zero-bias calibration.

[0140] Dual-frequency GNSS receivers achieve centimeter-level accuracy by receiving L1 / L2 satellite signals and supporting RTK differential positioning. The position and velocity information they output complement the inertial measurement data. Preferably, a tightly coupled navigation solution model processes GNSS pseudorange / Doppler observations and raw inertial data through a deep combination approach. Position information is used to constrain the cumulative error of inertial navigation, while velocity information assists in determining carrier phase cycle slips.

[0141] The improved robust UKF algorithm introduces an adaptive factor based on singular value decomposition (SVD) to the standard unscented Kalman filter framework to adjust the observation weight matrix. This algorithm automatically reduces the impact of individual, continuous, or mixed GNSS observation anomalies on state estimation. The position offset reflects the deviation between the actual motion trajectory of the unmanned vehicle 1 and the ideal trajectory, while the attitude correction includes the rotation error of the ship's coordinate system caused by wind and wave disturbances.

[0142] Preferably, the conversion relationship between the hull coordinate system and the world coordinate system is established through a homogeneous transformation matrix, and its rotation parameters are dynamically updated by real-time posture data. The conversion result is used to construct the dynamic activity space 4 of the unmanned boat 1.

[0143] Sonar measurement error compensation targets beam pointing offset and time delay measurement distortion caused by hull sway, and corrects the echo arrival time through kinematic inverse calculation.

[0144] Pure inertial navigation mode relies on MEMS inertial measurement unit (IMU) independent calculations during GNSS failures, and its duration is controlled by a threshold set based on inertial sensor accuracy. The navigation status flag is used to identify system operating mode switching events and includes status parameters such as GNSS availability and inertial navigation error level.

[0145] This embodiment can effectively compensate for the measurement errors caused by the motion state of the unmanned boat 1 through the tightly coupled navigation solution model of the MEMS inertial measurement unit and the dual-frequency GNSS receiver, combined with the data fusion processing of the improved anti-error UKF algorithm. This solution realizes the precise conversion of environmental detection information from the hull coordinate system to the world coordinate system, and significantly improves the accuracy of the environmental correction information of the obstacle position by calculating the position offset and attitude correction. When the GNSS signal is lost, the system can automatically switch to the pure inertial navigation mode and record the navigation status flag, ensuring the continuous reliability of the navigation compensation module. This embodiment effectively solves the problem of interference of hull motion on sonar measurement, and provides an accurate environmental correction information basis for the underwater obstacle avoidance of the unmanned boat 1.

[0146] In some embodiments, the safety domain construction module is used to construct the ellipsoidal space safety area of ​​the current unmanned vehicle based on the environmental correction information, including:

[0147] Obtain the obstacle positions in the environmental correction information one by one, and extract the azimuth and distance information of the obstacles relative to the unmanned boat;

[0148] The first actual position of the wide beam transmitter and the second actual position of the wide beam receiver at the current moment are used as dual focal points, and the ellipsoid parameters are calculated based on the sound wave propagation time difference, including:

[0149] The shortest propagation distance corresponding to the original environmental information of the earliest echo is taken as the major axis of the ellipsoid;

[0150] Determine the focal length of the ellipsoid according to the distance between the transmitting and receiving transducers, and calculate the minor axis of the ellipsoid according to the real-time draft depth of the unmanned boat;

[0151] Obtain the ellipsoid parameters according to the ellipsoid major axis, ellipsoid minor axis and ellipsoid focal length;

[0152] A dynamic ellipsoid safety model is constructed based on the ellipsoid parameters in the world coordinate system, including:

[0153] Establish the equation of the ellipsoid with the transmitting and receiving transducers as the dual foci;

[0154] Update the position in ellipsoid space according to real-time navigation data;

[0155] Visualize the safe area in ellipsoid space.

[0156] In this embodiment, environmental correction information refers to the obstacle's spatial position data processed by the navigation compensation module, including the obstacle's precise coordinates in the world coordinate system. The obstacle's azimuth and distance relative to the UUV 1 are calculated using the arrival time and beam pointing angle of the sonar echo signal. The azimuth represents the angle between the obstacle and the UUV's heading, while the distance reflects the linear separation between the obstacle and the vessel.

[0157] The first actual position of the wide-beam transmitter 21 and the second actual position of the wide-beam receiver 22 are the transducer coordinates after ship motion compensation. They are updated in real time by fusing GNSS positioning data with attitude data from the inertial measurement unit. The acoustic propagation time difference (TDOT) is the time interval between the transmitted signal and the earliest echo signal. This parameter is used to calculate the shortest propagation path of acoustic waves in water.

[0158] The ellipsoid's major axis is determined by the acoustic wave propagation distance corresponding to the original environmental information of the earliest echo, which is equal to half the major axis length. The spacing between the transceiver and transmitter serves as the basis for determining the ellipsoid's focal length, and its value is calculated based on the spatial geometry of the transducer installation locations.

[0159] The real-time draft depth of the unmanned boat 1 is obtained through a pressure sensor or a waterline detection device. This parameter is used to constrain the lower limit of the minor axis of the ellipsoid to ensure that the safety area does not include the hull structure itself.

[0160] The dynamic ellipsoid safety model is constructed by first establishing an ellipsoid equation based on a dual focus according to the principles of analytic geometry. The model's spatial position is then continuously updated using real-time navigation data. Preferably, the visualization of the ellipsoidal safety zone 3 is achieved through three-dimensional graphics rendering technology, with the display range dynamically adjusted based on the ellipsoid's parameters, providing the operator with intuitive spatial safety situational awareness. This embodiment uses a dynamic ellipsoid model to accurately characterize the safe waters surrounding the unmanned vehicle 1, providing a reliable spatial constraint basis for obstacle avoidance decisions.

[0161] This embodiment uses the first actual position of the wide-beam transmitter 21 and the second actual position of the wide-beam receiver 22 as dual foci to construct an ellipsoidal space safety zone 3. Based on the time difference in sound wave propagation, the ellipsoidal parameters are accurately calculated to achieve precise positioning of underwater obstacles. The real-time draft of the unmanned boat 1 is used to constrain the minor axis of the ellipsoid to ensure that the safety zone does not include the hull structure itself. The ellipsoidal space position is dynamically updated using real-time navigation data, and an ellipsoidal equation with the transceiver transducer as the dual foci is established to form a reliable spatial constraint basis. The visualization of the ellipsoidal space safety zone 3 intuitively displays the scope of the safe waters, providing effective support for obstacle avoidance decisions.

[0162] In some embodiments, the safe navigation boundary corresponding to the safe area in the ellipsoid space is calculated in real time to obtain safe navigation information, including:

[0163] Calculate the spatial relationship between the ellipsoidal space safety area and the unmanned boat's hull outline to generate the initial safe navigation boundary;

[0164] Outputting an initial safety area dataset of the ellipsoid space safety area according to the initial safety navigation boundary, the initial safety area dataset including an initial boundary coordinate set and an initial nearest obstacle warning level;

[0165] Based on the initial boundary coordinate set, a three-dimensional space grid map is established, where the three-dimensional space grid map includes a plurality of grid units;

[0166] Mark grid cells one by one as safe area, warning area or dangerous area according to the initial nearest obstacle warning level;

[0167] Obtain the dynamic parameters of the unmanned boat and calculate the navigable area based on the dynamic parameters. The dynamic parameters include the current speed, steering angle, roll angle, and pitch angle of the unmanned boat, including:

[0168] Generate the minimum turning radius according to the current speed and steering angle of the unmanned boat;

[0169] Generate the influence coefficient of the current minimum turning radius on the actual operable space based on the roll angle and pitch angle;

[0170] Generate the navigable area of ​​the three-dimensional grid map based on the minimum turning radius, influence coefficient and preset safety margin threshold;

[0171] Perform multi-level boundary analysis within the navigable domain, including:

[0172] Calculate the minimum safe distance from the ellipsoid surface to the nearest obstacle to obtain the primary boundary;

[0173] The track envelope within a preset short time interval is predicted based on the primary boundary and the current speed to generate the secondary boundary.

[0174] Generate dynamic safety corridors based on primary boundaries and secondary boundaries to obtain safe navigation boundaries;

[0175] Also, when it is detected that the width of the dynamic safety corridor is less than a preset multiple of the hull width, the speed reduction mode is automatically triggered and the safe navigation boundary is replanned;

[0176] Based on the safe navigation boundary, safe navigation information is generated, which includes the optimal safe navigation direction recommendation, the maximum allowable speed in each direction, the emergency braking distance parameters, and the multi-level warning status identification.

[0177] In this embodiment, the initial safe navigation boundary refers to the initial safety range generated by calculating the spatial geometric relationship between the ellipsoidal space safety area 3 and the hull outline of the unmanned boat 1. This boundary is used to define the basic spatial range in which the unmanned boat 1 can safely navigate.

[0178] The initial safe area dataset contains geometric information consisting of an initial boundary coordinate set and risk assessment information consisting of an initial nearest obstacle warning level. The initial boundary coordinate set is obtained through spatial geometric calculations, and the initial nearest obstacle warning level is determined by comprehensively determining the earliest arrival time and intensity of the sonar echo signal, and is specifically divided into three levels: safe, alert, and dangerous.

[0179] The construction of the three-dimensional spatial grid map is based on the world coordinate system. Each grid unit corresponds to the actual water space volume, and its size is determined according to the sonar detection accuracy and the size of the unmanned boat 1.

[0180] The dynamic parameters of the unmanned boat 1 are key parameters reflecting the motion characteristics of the unmanned boat 1, including the current speed, steering angle, roll angle and pitch angle. These parameters can be collected in real time through ship-borne sensors and used to calculate the actual operable space.

[0181] The minimum turning radius is calculated from the kinematic relationship between the current speed and steering angle of the UAV 1, representing the lower limit of the vessel's maneuverability under a specific motion state. The coefficient of influence of the current minimum turning radius on the actual maneuverable space is calculated using the UAV 1 dynamics model, which accounts for the nonlinear effects of roll and pitch angles on maneuverability.

[0182] The navigable area is the practically feasible region defined in a three-dimensional grid map based on the minimum turning radius, influence coefficient, and a preset safety margin threshold. This region takes into account the limitations of the ship's hull dynamics and environmental constraints. Preferably, the preset safety margin threshold is set based on the braking performance and operational response delay of the unmanned vehicle 1 to ensure that sufficient buffer space is left in the navigable area calculation.

[0183] The trajectory envelope prediction in the multi-level boundary analysis adopts a trajectory deduction algorithm based on kinematics, which fully considers the inertial characteristics and fluid dynamic effects of the unmanned boat 1.

[0184] The primary boundary is calculated by calculating the minimum safe distance from the ellipsoid surface to the nearest obstacle, reflecting the static safety margin at the current moment. The secondary boundary, based on the primary boundary and the current speed, predicts the trajectory envelope for a short period of time in the future and is used to assess dynamic navigation risks. A dynamic safety corridor is generated by integrating the primary and secondary boundaries. The criteria for determining the corridor are proportional to the hull width. When the corridor width is insufficient, the deceleration mode is triggered, which achieves controlled deceleration by gradually reducing propulsion power.

[0185] Safe navigation information is the final output basis for navigation decisions, and includes key parameters such as the optimal safe heading recommendation. These parameters are generated by comprehensively analyzing the spatial characteristics of the dynamic safety corridor and the hull motion constraints.

[0186] This embodiment provides accurate navigation decision support for the unmanned vehicle 1 through multi-level safety boundary calculation and dynamic risk assessment. Specifically, by collaboratively constructing an initial safe navigation boundary with a three-dimensional spatial grid map, the navigable area is accurately calculated in conjunction with the dynamic parameters of the unmanned vehicle 1. Multi-level boundary analysis is used to generate dynamic safety corridors. When insufficient corridor width is detected, the speed reduction mode is automatically triggered. The resulting safe navigation information provides the unmanned vehicle 1 with a comprehensive decision-making basis that balances static safety margins with dynamic navigation risks.

[0187] In some embodiments, the obstacle avoidance decision module is used to obtain the real-time navigation status and safe navigation information of the current unmanned vehicle and generate a multi-level obstacle avoidance strategy with priority, including:

[0188] Construct an obstacle avoidance decision tree based on risk assessment. The obstacle avoidance decision tree includes a first-level response decision layer, a second-level response decision layer, and a third-level response decision layer. The first-level response decision layer is configured to immediately trigger an emergency braking command when the collision time is less than 3 seconds. The second-level response decision layer is configured to generate a course correction strategy when the width of the dynamic safety corridor is less than 2 times the ship width. The third-level response decision layer is configured to select an obstacle avoidance route based on the principle of optimal energy consumption when there are multiple feasible paths.

[0189] The real-time navigation status and safe navigation information of the current unmanned boat are input into the obstacle avoidance decision tree to obtain the output results, and the output results are dynamically optimized, including:

[0190] Real-time calculation of the feasibility index of each obstacle avoidance solution in the output results. The feasibility index includes the path smoothness score, energy consumption cost score, and task delay cost score.

[0191] The feasibility index is weighted and integrated using fuzzy logic algorithm to obtain the comprehensive score of each obstacle avoidance solution;

[0192] The priorities of multiple obstacle avoidance schemes are generated based on the comprehensive scores, and a multi-level obstacle avoidance strategy package is generated. The multi-level obstacle avoidance strategy package is adaptively adjusted according to the control characteristics of the unmanned boat, including:

[0193] Add lateral displacement compensation for flat-bottomed ships;

[0194] Optimized turning rate parameters for V-shaped hull;

[0195] Dynamically adjust safety thresholds based on sea conditions;

[0196] When new obstacles appear during the execution of the strategy, the strategy re-planning mechanism is activated and the historical decision log is retained.

[0197] In this embodiment, the obstacle avoidance decision module is the core control unit for comprehensively processing the navigation status and safety information of the unmanned vehicle 1. It implements intelligent obstacle avoidance through a hierarchical decision-making mechanism. The risk-assessment-based obstacle avoidance decision tree adopts a three-level response architecture. The first level of the response decision-making layer implements emergency avoidance through real-time collision time determination, the second level of the response decision-making layer implements course correction based on the width of the dynamic safety corridor, and the third level of the response decision-making layer optimizes the path based on the principle of optimal energy consumption.

[0198] The real-time navigation status includes the current motion parameters and environmental perception data of the unmanned vehicle 1. These data, along with the safe navigation information, form the decision-making input. The feasibility index is derived by quantitatively evaluating the key performance indicators of the obstacle avoidance solution. The path smoothness score reflects the continuity of the track curvature, the energy cost score calculates the propulsion system power consumption, and the mission delay cost score assesses the travel time lost due to obstacle avoidance. A weighted fusion fuzzy logic algorithm uses a membership function to address the nonlinear relationships between the scores, ultimately outputting a comprehensive score that serves as the basis for optimal solution selection.

[0199] The generation of a multi-level obstacle avoidance strategy package takes into account the varying handling characteristics of different vessel types. Transverse displacement compensation is designed to address the drift characteristics of flat-bottomed vessels, while steering rate parameters are optimized to accommodate the turning performance of V-shaped hulls. A dynamic safety threshold adjustment mechanism uses real-time sea state data to update decision boundaries. A strategy replanning mechanism ensures rapid response to new obstacles, while historical decision logs provide data support for subsequent strategy optimization.

[0200] This embodiment combines a multi-level decision-making architecture with a dynamic optimization mechanism to achieve intelligent generation and real-time adjustment of obstacle avoidance strategies in complex environments. This system achieves multi-level response by constructing a risk-assessment-based obstacle avoidance decision tree. Fuzzy logic fusion with feasibility indexes generates a prioritized, multi-level obstacle avoidance strategy package, which is then adaptively adjusted to the characteristics of different ship types, ensuring real-time and reliable obstacle avoidance decision-making in complex environments.

[0201] In some embodiments, the execution control module is used to convert the multi-level obstacle avoidance strategy into obstacle avoidance instructions, where the obstacle avoidance instructions include heading control instructions and speed control instructions, including:

[0202] Convert the path planning parameters of each obstacle avoidance path in the multi-level obstacle avoidance strategy into heading control instructions, including:

[0203] Extract the heading angle sequence of the obstacle avoidance path;

[0204] A trajectory smoothing algorithm is used to generate continuous steering commands based on the heading angle sequence;

[0205] Calculate the actual rudder angle control value according to the rudder effect characteristics and obtain the heading control instruction;

[0206] Convert the speed planning parameters of each obstacle avoidance path in the multi-level obstacle avoidance strategy into speed control instructions, including:

[0207] Analyze the recommended speed under the safety distance constraint and generate the acceleration curve that meets the propulsion characteristics;

[0208] Calculate the final thrust control parameters according to the acceleration curve and obtain the speed control command;

[0209] Obstacle avoidance instructions are generated based on the heading control instructions and speed control instructions, and are output through the multi-mode output interface. The multi-mode output interface is configured as follows:

[0210] In normal cruise mode, the periodic command output mode is adopted;

[0211] Enable event-triggered command output mode in emergency obstacle avoidance mode;

[0212] Switch to degraded instruction output mode when the system is abnormal;

[0213] Also, safety verification is performed on obstacle avoidance instructions before output, including:

[0214] Verify the valid range of the command parameters of the obstacle avoidance command;

[0215] Limit the command change rate of obstacle avoidance commands;

[0216] Check the coordination between multiple instructions of obstacle avoidance instructions.

[0217] In this embodiment, the execution control module connects to the UAV's power control system via a control interface to achieve precise navigation control. Path planning parameters in the multi-level obstacle avoidance strategy are extracted through a heading angle sequence, which represents the set of key steering nodes in the obstacle avoidance path. A trajectory smoothing algorithm is used to eliminate sudden heading changes and generate continuous steering commands that conform to the vessel's maneuverability. Its smoothness is determined by the UAV's maneuverability.

[0218] The conversion process of speed planning parameters determines the base speed requirement through a safety distance constraint, which is dynamically adjusted based on the real-time calculation results of the ellipsoidal safety zone 3. The recommended speed is generated based on the current obstacle distribution, and the calculation process comprehensively considers the propulsion characteristics and braking performance of the unmanned vehicle 1. The operating mode switching conditions of the multi-mode output interface are directly related to the navigation status of the unmanned vehicle 1. The response time of the event-triggered command output mode must meet the real-time requirements of emergency obstacle avoidance.

[0219] The safety verification process ensures the reliability of the instructions through multi-dimensional verification. The effective range verification of the instruction parameters is set according to the physical limit of the power system of the unmanned boat 1, and the instruction change rate limit parameter is determined according to the dynamic response characteristics of the actuator.

[0220] This embodiment achieves high-fidelity translation from decision-making strategy to action execution through refined command conversion and multiple safety verification mechanisms. By precisely converting multi-level obstacle avoidance strategies into heading and speed control commands, combining trajectory smoothing algorithms with thrust control parameter calculations, and employing a multi-mode output interface and safety verification, the precise generation and reliable execution of obstacle avoidance commands are ensured.

[0221] In some embodiments, the output frequency of obstacle avoidance instructions is dynamically adjusted to adapt to different navigation conditions, including:

[0222] Determine the reference output frequency based on the current navigation conditions, including:

[0223] Obtain the real-time navigation status parameters of the unmanned boat;

[0224] Identify the current operating condition type based on real-time navigation status parameters;

[0225] Query the working condition-frequency mapping table to determine the reference output frequency;

[0226] When the speed of the real-time navigation status parameter increases, the reference output frequency is gradually increased, and the frequency adjustment log information is recorded;

[0227] Perform smooth transition processing during reference output frequency adjustment, including:

[0228] Set the frequency switching transition interval and adopt a gradual adjustment algorithm to maintain the continuity of the obstacle avoidance command output process.

[0229] In this embodiment, dynamically adjusting the output frequency of obstacle avoidance commands is a key mechanism for adaptively adjusting the control command update rate based on changes in the actual navigation conditions of the unmanned boat 1. The current navigation condition is identified by analyzing real-time navigation state parameters, including but not limited to key indicators such as speed and steering angle that characterize the vessel's motion state. The operating condition type refers to different operating modes divided according to the navigation state. Preferably, these are specifically divided into four typical scenarios: low-speed cruising, conventional navigation, high-speed navigation, and emergency obstacle avoidance. This division is based primarily on the differentiated requirements for control response timeliness in different speed ranges.

[0230] The operating condition-frequency mapping table stores preset baseline control frequencies corresponding to different operating conditions. This mapping relationship was determined through dynamic characteristic testing of the unmanned vehicle 1 to ensure optimal alignment between the control command update frequency and the motion state under each operating condition. The baseline output frequency serves as the reference for regulation, and a frequency adjustment log records each adjustment process, facilitating system tracing and analysis of frequency change patterns. When a speed increase is triggered, the system gradually increases the output frequency according to preset rules to ensure synchronization of control commands and speed changes.

[0231] Smooth transition processing achieves bumpless switching through a frequency switching transition interval. Optionally, the frequency switching transition interval can be set to a duration of 20%-30% of the difference between the current and target frequencies, ensuring that the actuator can smoothly adapt to frequency changes. The progressive adjustment algorithm uses an S-curve transition function to achieve continuity in frequency changes. The instantaneous frequency value in the transition interval is calculated through nonlinear interpolation to avoid control jitter caused by sudden frequency changes.

[0232] This embodiment utilizes a dynamic frequency adjustment mechanism to ensure that the output frequency of obstacle avoidance commands is always optimally matched to the current navigation conditions. By dynamically adjusting the output frequency of obstacle avoidance commands, the control command update rate is consistently adapted to the current navigation conditions. Combined with smooth transition processing to ensure command continuity, this effectively improves the timely obstacle avoidance response and control stability of the unmanned vehicle 1 under various navigation conditions.

[0233] In some embodiments, the unmanned boat underwater obstacle avoidance system based on distributed wide-beam sonar also includes a reverberation suppression module and a zero-range blind zone compensation module. The reverberation suppression module is used to establish a seabed-water surface multipath reflection channel model, use an adaptive notch filter to suppress reverberation in a specific frequency band, eliminate false echoes, and adjust the signal pulse width of the transmitted signal to optimize the distance resolution; the zero-range blind zone compensation module is used to establish an error compensation model based on hull vibration, integrate inertial data to predict the position of proximal obstacles, and use a time-gated algorithm to isolate self-interference signals.

[0234] In this embodiment, the reverberation suppression module analyzes the characteristics of acoustic wave reflection paths by establishing a seabed-surface multipath reflection channel model. This model includes propagation characteristic parameters for direct waves, surface reflections, and seabed reflections. An adaptive notch filter suppresses reverberation in specific frequency bands by dynamically adjusting its filtering characteristics. Its filtering parameters are automatically updated based on the real-time monitored reverberation signal spectrum. Signal pulse width adjustment is implemented through a pulse generation circuit, dynamically adjusting the transmit pulse width based on current detection requirements to optimize range resolution.

[0235] The zero-range blind zone compensation module establishes an error compensation model based on hull vibration. This model integrates attitude data output by the inertial measurement unit to predict the displacement trajectory of obstacles near the hull. A time-gating algorithm precisely controls the signal reception period to isolate self-interference signals. Its gating timing is synchronized with the transmitted pulse to ensure accurate extraction of valid echo signals. By fusing sonar measurement data with inertial navigation data, the module reliably predicts the position of obstacles within the zero-range blind zone.

[0236] This embodiment effectively reduces multipath reflection interference through the adaptive notch filter and signal pulse width adjustment of the reverberation suppression module. Combined with the hull vibration error compensation model and time gating algorithm of the zero-range blind zone compensation module, it significantly improves the near-end obstacle detection accuracy and system reliability.

[0237] By adopting the above technical solution, the present invention has the following beneficial effects compared with the prior art:

[0238] The present invention effectively improves the reliability of underwater environmental detection by leveraging the non-directional, wide-angle detection characteristics of the distributed wide-beam sonar array 2, in conjunction with adaptive filtering. The sonar detection module utilizes a split design featuring a wide-beam transmitter 21 and a wide-beam receiver 22. This design incorporates asynchronous co-frequency or synchronous off-frequency modulation signal transmission strategies, and employs dynamic noise reduction through adaptive filtering based on the NLMS algorithm to generate environmental detection information with an improved signal-to-noise ratio. The navigation compensation module utilizes a tightly coupled system constructed from a MEMS-IMU and a dual-frequency GNSS receiver to acquire the motion state of the unmanned vehicle 1. Data fusion employs an improved robust UKF algorithm based on SVD, compensating for motion errors in the environmental detection information to generate environmental correction information. The safety domain construction module constructs an ellipsoidal safety zone 3 for the current unmanned vehicle 1 based on this environmental correction information, calculating the safe navigation boundary in real time to obtain safe navigation information. The obstacle avoidance decision module acquires real-time navigation status and safe navigation information, generating a prioritized, multi-level obstacle avoidance strategy. The execution control module converts multi-level obstacle avoidance strategies into obstacle avoidance commands, including heading and speed control commands, and dynamically adjusts the output frequency to adapt to different navigation conditions. The reverberation suppression module uses an adaptive notch filter to suppress reverberation in specific frequency bands, and the zero-range blind zone compensation module establishes an error compensation model based on hull vibration, jointly improving the accuracy of near-end obstacle detection. These modules are coordinated and controlled by a processing unit based on an industrial-grade MCU, ultimately achieving real-time matching between the unmanned vehicle's 1 dynamic activity space 4 and the ellipsoidal space safety zone 3.

[0239] In addition, the functional units in various embodiments of the present invention may be integrated into a single processing unit, each unit may exist physically separately, or two or more units may be integrated into a single unit. The aforementioned integrated units may be implemented in the form of hardware or software functional units.

[0240] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) or a processor to execute all or part of the steps of the various embodiments of the present invention. The aforementioned storage medium includes various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0241] The above descriptions are only some embodiments of the present invention and do not limit the scope of protection of the present invention. Any equivalent device or equivalent process transformation made by using the contents of the description and drawings of the present invention, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. An underwater obstacle avoidance system for an unmanned boat based on distributed wide-beam sonar, characterized in that: include: A sonar detection module, comprising a wide-beam sonar group, each of which comprises a wide-beam transmitter and a wide-beam receiver. The sonar detection module is used to collect raw environmental information and dynamically reduce noise on the raw environmental information using adaptive filtering to obtain environmental detection information. A navigation compensation module is used to obtain the motion state of the unmanned boat and perform motion error compensation on the environmental detection information to obtain environmental correction information; The safety domain construction module is used to construct the ellipsoidal space safety area of ​​the current unmanned vehicle according to the environmental correction information, including: Obtaining the positions of obstacles in the environmental correction information one by one, and extracting the azimuth and distance information of the obstacles relative to the unmanned boat; The first actual position of the wide beam transmitter and the second actual position of the wide beam receiver at the current moment are used as dual focal points, and the ellipsoid parameters are calculated based on the sound wave propagation time difference, including: The shortest propagation distance corresponding to the original environmental information of the earliest echo is taken as the major axis of the ellipsoid; Determine the focal length of the ellipsoid according to the distance between the transmitting and receiving transducers, and calculate the minor axis of the ellipsoid according to the real-time draft depth of the unmanned boat; Obtain the ellipsoid parameters according to the ellipsoid major axis, the ellipsoid minor axis and the ellipsoid focal length; Constructing a dynamic ellipsoid safety model according to the ellipsoid parameters in a world coordinate system, including: Establish the equation of the ellipsoid with the transmitting and receiving transducers as the dual foci; Update the position in ellipsoid space according to real-time navigation data; Visualizing the ellipsoidal space safety area; The safe navigation boundary corresponding to the safe area in the ellipsoid space is calculated in real time to obtain safe navigation information, including: Calculating the spatial relationship between the ellipsoidal space safety area and the hull outline of the unmanned boat to generate an initial safe navigation boundary; Outputting an initial safety area data set of the ellipsoid space safety area according to the initial safety navigation boundary, the initial safety area data set including an initial boundary coordinate set and an initial nearest obstacle warning level; Establishing a three-dimensional space grid map based on the initial boundary coordinate set, wherein the three-dimensional space grid map includes a plurality of grid cells; Marking grid cells one by one as safe areas, warning areas or dangerous areas according to the initial nearest obstacle warning level; Obtain the dynamic parameters of the unmanned boat and calculate the navigable area based on the dynamic parameters, wherein the dynamic parameters include the current speed, steering angle, roll angle, and pitch angle of the unmanned boat, including: Generate the minimum turning radius according to the current speed and steering angle of the unmanned boat; Generate the influence coefficient of the current minimum turning radius on the actual operable space based on the roll angle and pitch angle; Generate a navigable area of ​​a three-dimensional space grid map according to the minimum turning radius, the influence coefficient and a preset safety margin threshold; The obstacle avoidance decision module is used to obtain the real-time navigation status and safe navigation information of the current unmanned boat and generate a multi-level obstacle avoidance strategy with priority; The execution control module is used to convert the multi-level obstacle avoidance strategy into obstacle avoidance instructions, which include heading control instructions and speed control instructions, and dynamically adjust the output frequency of the obstacle avoidance instructions to adapt to different navigation conditions.

2. The unmanned boat underwater obstacle avoidance system based on distributed wide-beam sonar according to claim 1 is characterized in that: The sonar detection module is used to collect original environmental information and use adaptive filtering to dynamically reduce noise on the original environmental information to obtain environmental detection information, including: Controlling the wide-beam transmitter to transmit an acoustic signal, wherein the wide-beam transmitter adopts linear frequency modulation, and the transmission power of the wide-beam transmitter is generated according to the reverberation intensity monitored in real time; The underwater echo signal is received by the wide beam receiver, and the underwater echo signal is amplified, band-pass filtered, and analog-to-digital converted to obtain original environmental information; Adaptive filtering is performed on the original environmental information, including: Performing primary filtering on the original environmental information using a normalized least mean square algorithm to obtain first filtered information; Performing secondary filtering on the first filtered information using a convolutional neural network model to obtain second filtered information; Using a pulse compression algorithm to perform time delay estimation on the second filtered information, correcting the earliest echo arrival time, and obtaining third filtered information; Calculating the obstacle distance based on the sound speed and the third filter information, and generating environment detection information including the obstacle position, obstacle distance, and obstacle strength; Also, in the event of communication anomalies, the environmental detection information is temporarily stored in the local cache, and the data collection timestamp and signal quality indicators are recorded.

3. The unmanned boat underwater obstacle avoidance system based on distributed wide-beam sonar according to claim 2 is characterized in that: The convolutional neural network model is trained by the following steps: Constructing an initial network model and inputting sample data for training, wherein the sample data includes a sample echo signal, and the initial network model includes a feature extraction layer and a feature classification layer, wherein the feature extraction layer is used to perform multi-scale feature extraction on the sample echo signal to obtain a sample target echo and a sample noise component, and the feature classification layer is used to distinguish the sample target echo from the sample noise component; The initial network model is trained repeatedly until the classification accuracy reaches a preset threshold, thereby obtaining the convolutional neural network model.

4. The unmanned boat underwater obstacle avoidance system based on distributed wide-beam sonar according to claim 1, characterized in that: The navigation compensation module is used to obtain the motion state of the unmanned boat and perform motion error compensation on the environmental detection information to obtain environmental correction information, including: The real-time attitude data of the unmanned boat is collected by a MEMS inertial measurement unit, wherein the real-time attitude data includes roll angle, pitch angle and heading angle; Acquire the position information and speed information of the unmanned boat through a dual-frequency GNSS receiver, and construct a tightly coupled navigation solution model based on the position information, speed information and the real-time attitude data; Performing data fusion on the tightly coupled navigation solution model based on the improved robust UKF algorithm to calculate the motion compensation parameters of the unmanned boat in the world coordinate system, wherein the motion compensation parameters include position offset and attitude correction; Performing coordinate transformation compensation on the environment detection information according to the motion compensation parameter includes: Establish the conversion relationship between the hull coordinate system and the world coordinate system; Convert the obstacle distance in the environment detection information from the ship coordinate system to the world coordinate system; Compensate for sonar measurement errors caused by ship motion; Outputs environmental correction information containing the precise position of obstacles in the world coordinate system; Also, when the GNSS signal is lost, it switches to pure inertial navigation mode and records the navigation status flag.

5. The unmanned boat underwater obstacle avoidance system based on distributed wide-beam sonar according to claim 1 is characterized in that: Calculate the safe navigation boundary corresponding to the safe area in the ellipsoid space in real time to obtain safe navigation information, including: Perform multi-level boundary analysis within the navigable area, including: Calculate the minimum safe distance from the ellipsoid surface to the nearest obstacle to obtain the primary boundary; Generate a secondary boundary by predicting a track envelope within a preset short time interval in the future based on the primary boundary and the current speed; generating a dynamic safety corridor according to the primary boundary and the secondary boundary to obtain the safe navigation boundary; Also, when it is detected that the width of the dynamic safety corridor is less than a preset multiple of the hull width, the speed reduction mode is automatically triggered and the safe navigation boundary is replanned; Safe navigation information is generated based on the safe navigation boundary, and the safe navigation information includes an optimal safe navigation direction suggestion, a maximum allowable speed in each direction, an emergency braking distance parameter, and a multi-level warning status indicator.

6. The unmanned boat underwater obstacle avoidance system based on distributed wide-beam sonar according to claim 1, characterized in that: The obstacle avoidance decision module is used to obtain the real-time navigation status and safe navigation information of the current unmanned boat and generate a multi-level obstacle avoidance strategy with priority, including: Constructing a risk assessment-based obstacle avoidance decision tree, the obstacle avoidance decision tree includes a primary response decision layer, a secondary response decision layer, and a tertiary response decision layer. The primary response decision layer is configured to immediately trigger an emergency braking command when the collision time is detected to be less than 3 seconds. The secondary response decision layer is configured to generate a course correction strategy when the dynamic safety corridor width is less than 2 times the ship width. The tertiary response decision layer is configured to select an obstacle avoidance route based on the principle of optimal energy consumption when multiple feasible paths exist. Inputting the current real-time navigation status and safe navigation information of the unmanned boat into the obstacle avoidance decision tree to obtain an output result, and performing dynamic strategy optimization on the output result, including: Calculate the feasibility index of each obstacle avoidance solution in the output results in real time, including the path smoothness score, energy consumption cost score, and task delay cost score; The feasibility index is weighted and integrated using a fuzzy logic algorithm to obtain a comprehensive score for each obstacle avoidance solution; The priorities of multiple obstacle avoidance schemes are generated according to the comprehensive scoring scores, and a multi-level obstacle avoidance strategy package is generated.

7. The unmanned boat underwater obstacle avoidance system based on distributed wide-beam sonar according to claim 1, characterized in that: The execution control module is used to convert the multi-level obstacle avoidance strategy into an obstacle avoidance instruction, wherein the obstacle avoidance instruction includes a heading control instruction and a speed control instruction, including: Converting the path planning parameters of each obstacle avoidance path in the multi-level obstacle avoidance strategy into a heading control instruction includes: Extract the heading angle sequence of the obstacle avoidance path; generating continuous steering instructions according to the heading angle sequence using a trajectory smoothing algorithm; Calculate the actual rudder angle control value according to the rudder effect characteristics to obtain the heading control instruction; Converting the speed planning parameters of each obstacle avoidance path in the multi-level obstacle avoidance strategy into a speed control instruction includes: Analyze the recommended speed under the safety distance constraint and generate the acceleration curve that meets the propulsion characteristics; Calculating the final thrust control parameter according to the acceleration curve to obtain the speed control instruction; An obstacle avoidance instruction is generated according to the heading control instruction and the speed control instruction, and is output through a multi-mode output interface, wherein the multi-mode output interface is configured as follows: In normal cruise mode, the periodic command output mode is adopted; Enable event-triggered command output mode in emergency obstacle avoidance mode; Switch to degraded instruction output mode when the system is abnormal; Also, safety verification is performed on obstacle avoidance instructions before output, including: Verify the valid range of the command parameters of the obstacle avoidance command; Limit the command change rate of obstacle avoidance commands; Check the coordination between multiple instructions of obstacle avoidance instructions.

8. The unmanned boat underwater obstacle avoidance system based on distributed wide-beam sonar according to claim 7, characterized in that: Dynamically adjust the output frequency of obstacle avoidance commands to adapt to different navigation conditions, including: Determine the reference output frequency based on the current navigation conditions, including: Obtain the real-time navigation status parameters of the unmanned boat; Identifying the current operating condition type according to the real-time navigation status parameter; Query the working condition-frequency mapping table to determine the reference output frequency; When the speed of the real-time navigation status parameter increases, the reference output frequency is gradually increased, and the frequency adjustment log information is recorded; Performing a smooth transition process during the reference output frequency adjustment process includes: Set the frequency switching transition interval and adopt a gradual adjustment algorithm to maintain the continuity of the obstacle avoidance command output process.

9. The unmanned boat underwater obstacle avoidance system based on distributed wide-beam sonar according to claim 1, characterized in that: Also includes: The reverberation suppression module is used to establish a seabed-surface multipath reflection channel model, using an adaptive notch filter to suppress reverberation in specific frequency bands, eliminate false echoes, and adjust the pulse width of the transmitted signal to optimize distance resolution; The zero-range blind zone compensation module is used to establish an error compensation model based on hull vibration, fuse inertial data to predict the position of proximal obstacles, and use a time-gated algorithm to isolate self-interference signals.

Images