Underwater topographic information multi-mode sensing system and method

Abstract

The invention discloses an underwater topographic information multi-modal sensing system and method, and belongs to the field of topographic information sensing systems.A phased array ultrasonic ranging unit, a miniature water pressure sensor array and a nine-axis inertial measurement unit are integrated through a multi-modal sensing module, and a non-vision dominant underwater sensing system is constructed. The tactile feedback module solves the problem that a traditional local vibration module is fuzzy in orientation prompt through the space coding rule of a distributed vibration device, the multi-channel information transmission module adopts the polarization imaging and underwater acoustic communication fusion technology, multi-mode data is modulated to four polarization states through Stokes parameters, and therefore the orientation prompt of the multi-mode data is more accurate. By combining the orthogonal frequency division multiplexing technology of the OFDM underwater acoustic transmitter, the problems of limited bandwidth and inconvenience in real-time transmission of high-dimensional data in the traditional underwater acoustic communication are solved, the data interaction reliability among the modules is high, the hardware architecture is compact, the expandability is high, and the application prospect is wide.

Description

Technical Field

[0001] The present invention relates to the field of terrain information perception systems, and more specifically, to a multimodal perception system and method for underwater terrain information. Background Art

[0002] Underwater terrain perception technology has important application value in diving operations, ocean exploration, and special environment rescue. Traditional underwater perception systems mostly rely on vision-dominated detection methods, such as optical cameras, lidar, etc., but are limited by the complex physical environment underwater. Factors such as the attenuation effect of light in water, scattering of suspended particles, and water turbidity will cause the visual perception range to be significantly reduced. Especially in low-visibility or lightless environments (such as the deep sea, inside a sunken ship, or at a disaster site), vision-dominated systems are difficult to work effectively, which seriously restricts the environmental perception capabilities of divers or underwater equipment. Although existing underwater obstacle avoidance solutions have introduced non-visual perception methods such as sonar and water pressure sensors, the information presentation method is still mainly two-dimensional screen display or local tactile feedback, which has the following limitations:

[0003] 1. Single perception dimension: Traditional systems often only provide single-modal data on obstacle distance or direction, lacking comprehensive perception and fusion processing of multi-dimensional environmental information such as terrain slope and diver posture, making it difficult to establish a complete understanding of underwater terrain.

[0004] 2. Inefficient human-computer interaction: Existing tactile feedback devices mostly use localized vibration modules (such as a single area on the hand or waist), making it difficult to accurately indicate the location of obstacles through spatial coding. For example, when an obstacle is located in the left front or right rear, traditional devices cannot guide users to quickly determine the spatial location through differentiated feedback, resulting in delayed obstacle avoidance response.

[0005] 3. Data transmission limitations: Underwater wireless communications face challenges such as limited underwater acoustic channel bandwidth and significant multipath effects. Traditional underwater acoustic communication technology is not convenient for real-time transmission of high-dimensional environmental data (such as three-dimensional terrain coordinates and multimodal perception parameters), which restricts the efficiency of building remote three-dimensional terrain models.

[0006] In response to the above problems, there is an urgent need for an underwater terrain perception system that breaks through the limitations of "vision dominance". Therefore, we propose a multimodal perception system and method for underwater terrain information to solve the above problems. Summary of the Invention

[0007] 1. Technical problems to be solved

[0008] In response to the problems existing in the prior art, the purpose of the present invention is to provide a multimodal perception system and method for underwater terrain information. It provides a new technical path to solve the above problems by designing a full-body distributed vibration device, constructing a terrain vibration mapping model, and introducing polarization imaging and underwater acoustic communication fusion technology through a multimodal perception system for underwater terrain information enhanced by tactile feedback.

[0009] 2. Technical solution

[0010] To solve the above problems, the present invention adopts the following technical solutions.

[0011] A multimodal perception system for underwater terrain information, comprising a multimodal perception module, a data processing module, a tactile feedback module, and a multi-channel information transmission module;

[0012] The multimodal sensing module is connected to the input port of the data processing module via an SPI bus for data communication. The multimodal sensing module includes a phased array ultrasonic ranging unit, a micro water pressure sensor array, and a nine-axis inertial measurement unit, and is used to collect multimodal data such as obstacle distance, terrain slope, and diver's posture.

[0013] The data processing module includes an embedded processor and a storage unit. The data processing module is connected to the input port of the tactile feedback module through an SPI bus for data communication. The tactile feedback module interacts with the multi-channel information transmission module in real time through the UWM1000 underwater acoustic communication protocol stack to transmit the execution status data of the vibration control signal. The data processing module also includes a data fusion unit, a pattern generation unit and a pre-stored terrain-vibration mapping database. The data fusion unit uses an adaptive Kalman filter algorithm to perform spatiotemporal calibration on the received multimodal data. The pattern generation unit generates a vibration control signal containing quantitative parameters of obstacle azimuth α, distance d and terrain slope θ according to the pre-stored terrain-vibration mapping database. The signal parameters include 20-200Hz frequency, 0.1-5m / s 2 intensity, 0-π phase difference;

[0014] The tactile feedback module is connected to a vibration device via a distributed I2C bus. The vibration device includes a chest vibrator, a waist vibrator, and a limb vibration array. Each vibration device has a built-in micro vibration motor and a drive circuit. According to the control signal output by the data processing module, the spatial distribution of the vibration devices in different areas and the vibration phase difference are used to achieve spatial coding prompts of the obstacle orientation.

[0015] The multi-channel information transmission module includes a liquid crystal polarization modulator, an OFDM underwater acoustic transmitter, an electro-optical conversion module and a transducer. The transducer is connected to an external terminal via an SPI bus. The electro-optical conversion module encodes the processed multimodal data into an optical signal containing polarization degree and polarization angle information, which is converted by the transducer into an underwater acoustic signal and transmitted to the external terminal. The external terminal uses a voxel grid algorithm to construct a three-dimensional underwater terrain model based on the received signal.

[0016] Furthermore, the phased array ultrasonic ranging unit includes a 16-channel ultrasonic transducer array, uses a delay-add beamforming algorithm to form a ±60° fan-shaped detection area, and cooperates with the time-of-flight method to calculate obstacle distance data. The delay-add beamforming algorithm realizes beam direction control through the delay time difference of each channel ultrasonic transducer. The micro water pressure sensor array includes 8 groups of MEMS piezoresistive sensors, and the 8 groups of MEMS piezoresistive sensors are distributed in a two-dimensional matrix with a spacing of 10 mm on the detection surface. The terrain slope is solved by a differential pressure algorithm. The differential pressure algorithm calculates the terrain gradient through the pressure difference of adjacent sensors.

[0017] Furthermore, the data fusion unit establishes a water flow disturbance model and a state space model. The water flow disturbance model is a first-order Gauss-Markov process that describes the random interference of water flow velocity on sensor measurement. The state vector of the state space model includes three-dimensional coordinates, slope θ and sensor deviation terms. The observation equation of the state space model is z k =Hx k +v k ,

[0018] Where H is the observation matrix, v k is zero-mean Gaussian white noise with a covariance matrix of R. The time synchronization and error correction of multi-source data are achieved through extended Kalman filtering, and the fused three-dimensional terrain coordinate data in the format of (X, Y, Z, θ) is output, where θ represents the heading angle parameter calculated based on the nine-axis inertial measurement unit.

[0019] Furthermore, the vibration pattern of the tactile feedback module includes a spatial coding rule, which is: the chest vibrator corresponds to the 0-180° area in front, the waist vibrator corresponds to the 45-135° area on the left / right, and the limb vibration arrays are distributed in four quadrants: the front of the arm, the back of the arm, the front of the leg, and the back of the leg. The left front of the arm corresponds to 90° to the left, the right front of the arm corresponds to 270° to the right, the left front of the leg corresponds to 225° to the lower left, and the right front of the leg corresponds to 315° to the lower right. The obstacle orientation information is transmitted through the phase difference of 0-π of the vibrators in different quadrants. When the phase difference is 0, it corresponds to 0° in front, when the phase difference is π, it corresponds to 180° on the back, and when the phase difference is π / 2, it corresponds to 90° to the left or 270° to the right.

[0020] Furthermore, the multi-channel information transmission module adopts polarization imaging and underwater acoustic communication fusion technology, supports QPSK modulation, and modulates the terrain data to four polarization states of 45°, 135°, 90°, and 0° through the Stokes parameter. Among them, the Stokes parameter S1 corresponds to the 45° polarization state, that is, the obstacle distance, S2 corresponds to the 135° polarization state, that is, the terrain slope, S0 corresponds to the 90° polarization state, and S3 corresponds to the 0° polarization state, that is, the attitude data component, and uses orthogonal frequency division multiplexing technology to realize underwater data transmission.

[0021] A multimodal perception method for underwater terrain information comprises the following steps:

[0022] S1. Data Acquisition and Synchronization: The multimodal perception module is based on the PTP precision time protocol. Through the PTP master node integrated into the embedded processor and connected to the PTP modules of each sensor slave node via independent synchronization control lines, it synchronizes the phased array ultrasonic ranging unit, micro water pressure sensor array, and nine-axis inertial measurement unit at the nanosecond level based on the IEEE 1588 protocol. It also synchronously collects distance, slope, and attitude data, and transmits them to the data processing module via the SPI bus.

[0023] S2. Data Fusion and Signal Generation: The data processing module performs spatiotemporal registration of multi-source data through the data fusion unit, eliminates water flow noise and sensor drift using the adaptive Kalman filter algorithm, and generates a vibration control signal containing spatial positioning information by matching the terrain-vibration mapping database through the pattern generation unit.

[0024] S3. Information transmission: The tactile feedback module drives the vibration device at the corresponding part according to the control signal, and transmits obstacle distance, direction and terrain slope information to the user through a combination of differentiated vibration frequency, intensity and phase difference;

[0025] S4. Polarization imaging coding: The multi-channel information transmission module encodes the fused terrain data into a polarization image signal through a liquid crystal polarization modulator. After modulation by an OFDM underwater acoustic transmitter, it is transmitted to an external terminal through a transducer. The external terminal constructs a three-dimensional terrain model based on polarization demodulation and a voxel grid algorithm. The voxel grid algorithm is based on the discretization of three-dimensional space and characterizes terrain features through the occupancy rate of voxel units.

[0026] Furthermore, the polarization imaging encoding process described in step S4 is as follows: by mapping the obstacle distance to the polarization degree, mapping the terrain slope to the polarization angle, and mapping the attitude data to the Stokes vector component through the rotation matrix transformation from the Euler angle to the Cartesian coordinate system, a polarization composite signal containing multimodal information is formed, which is loaded onto the OFDM subcarrier for transmission after QPSK modulation. The rotation matrix transformation formula is: R(α, β, γ) = R z (γ)R y (β)R x (α).

[0027] Furthermore, the voxel grid algorithm in step S4 is based on three-dimensional space discretization, the voxel unit size is Δx×Δy×Δz=0.1m×0.1m×0.1m, the occupancy threshold is set to 0.6, and the terrain characteristics are characterized by the occupancy of the voxel unit.

[0028] 3. Beneficial effects

[0029] Compared with the prior art, the advantages of the present invention are:

[0030] (1) This solution integrates a phased array ultrasonic ranging unit, a micro water pressure sensor array, and a nine-axis inertial measurement unit through a multi-modal perception module to build a non-vision-dominated underwater perception system. The 16-channel ultrasonic transducer array uses a delay-addition beamforming algorithm to form a ±60° fan-shaped detection area, and cooperates with the time-of-flight method to achieve obstacle distance measurement, avoiding the constraints of underwater light attenuation on the visual system. It is particularly suitable for low-visibility environments such as deep sea and inside sunken ships. 8 groups of MEMS piezoresistive sensors distributed in a two-dimensional matrix calculate the terrain slope through a differential pressure algorithm. Combined with the attitude data obtained by the nine-axis inertial measurement unit, it realizes the simultaneous collection of multi-dimensional environmental information such as obstacle distance, terrain slope, and diver attitude, breaking through the limitation of traditional systems that only provide single-modal data, and constructing a complete underwater terrain cognition model including three-dimensional coordinates (X, Y, Z) and slope θ.

[0031] (2) In this solution, the tactile feedback module solves the problem of fuzzy orientation prompts in traditional local vibration modules through the spatial encoding rules of distributed vibration devices. The chest vibrator, waist vibrator and limb vibration array divide the space into front, left / right, and four quadrants, and encode the obstacle azimuth α through the 0-π phase difference (such as the phase difference 0 corresponds to 0° in front, π corresponds to 180° behind), combined with 20-200Hz frequency, 0.1-5m / s 2 Strength parameters enable multi-dimensional tactile encoding of obstacle orientation, distance, and slope, improving the real-time performance and safety of underwater operations.

[0032] (3) In this solution, the multi-channel information transmission module adopts the fusion technology of polarization imaging and underwater acoustic communication. The multi-modal data is modulated into four polarization states of 45°, 135°, 90°, and 0° through the Stokes parameter. Combined with the orthogonal frequency division multiplexing technology of the OFDM underwater acoustic transmitter, it solves the problem of limited bandwidth of traditional underwater acoustic communication and the inconvenience of real-time transmission of high-dimensional data.

[0033] (4) In this solution, low-power, high-reliability data interaction is achieved between modules through the SPI bus, distributed I2C bus, and UWM1000 underwater acoustic communication protocol stack. The hardware architecture is compact and highly scalable. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 This is a schematic diagram of the main system architecture of the present invention;

[0035] Figure 2 This is a schematic diagram of the composition and principle of the multimodal perception module of the present invention;

[0036] Figure 3 This is a schematic diagram of the composition and principle of the data processing module of the present invention;

[0037] Figure 4 This is a schematic diagram of the principle of spatial encoding rules of the tactile feedback module of the present invention;

[0038] Figure 5 This is a schematic diagram showing the principle of data transmission mode of the multi-channel information transmission module of the present invention;

[0039] Figure 6 The figure is a schematic diagram of the steps and principles of the method of the present invention.

[0040] Description of the numbers in the figure:

[0041] 100. Multimodal sensing module; 101. Phased array ultrasonic ranging unit; 1011. Ultrasonic transducer array; 102. Micro water pressure sensor array; 1021. MEMS piezoresistive sensor; 103. Nine-axis inertial measurement unit;

[0042] 200, data processing module; 201, embedded processor; 202, storage unit; 203, data fusion unit; 2031, water flow disturbance model; 2032, state space model; 204, pattern generation unit; 205, terrain-vibration mapping database;

[0043] 300, tactile feedback module; 301, vibration device; 302, chest vibrator; 303, waist vibrator; 304, limb vibration array;

[0044] 400, multi-channel information transmission module; 401, liquid crystal polarization modulator; 402, OFDM underwater acoustic transmitter; 403, electro-optical conversion module; 404, transducer;

[0045] 500. External terminal. DETAILED DESCRIPTION

[0046] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the specification of the present invention; it is obvious that the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0047] Example 1:

[0048] See also Figures 1-6 The following is a detailed description of the working principle of the underwater terrain information multimodal perception system and method based on the technical solution content and combined with technical details:

[0049] 1. Multimodal Data Spatiotemporal Synchronous Acquisition Mechanism

[0050] The multimodal perception module 100 achieves three-dimensional capture of environmental information through a heterogeneous sensor array and a precise synchronization protocol:

[0051] 1. The 16-channel ultrasonic transducer array 1011 of the phased array ultrasonic ranging unit 101 uses a delay-and-add beamforming algorithm. The FPGA controls the transmission delay difference of each channel to form a dynamically focused beam within a ±60° fan-shaped area. For example, when detecting an obstacle in the left front, the left channel triggers ultrasonic transmission in advance, deflecting the beam to the left. Time of flight (ToF) is used to calculate the echo time difference and determine the obstacle's distance.

[0052] 2. The eight MEMS piezoresistive sensors 1021 in the micro-water pressure sensor array 102 are distributed across the detection surface in a 2×4 two-dimensional matrix with a 10mm pitch. The pressure gradient between adjacent sensors is calculated using a differential pressure algorithm (ΔP = P1 - P2). This is combined with a fluid dynamics model to invert the terrain slope θ (resolution 0.1°). For example, as the sensor array moves along a slope, the pressure difference between the upstream and downstream sensors increases linearly. This is then fitted with a polynomial to output the slope parameter.

[0053] 3. The nine-axis inertial measurement unit 103 outputs Euler angles (heading angle ψ, pitch angle θ, roll angle φ) in real time through the built-in gyroscope, accelerometer and magnetometer, and combines with the extended Kalman filter to eliminate attitude drift (noise standard deviation <0.5° / h). Time synchronization core technology:

[0054] Based on the IEEE1588PTP protocol, the embedded processor 201 acts as the PTP master node and sends timestamps to each sensor slave node through an independent synchronization control line to achieve nanosecond clock alignment. After synchronization, the three types of sensors synchronously collect data at a frequency of 100 Hz and transmit the data to the data processing module 200 via the SPI bus to ensure the temporal and spatial consistency of distance, slope, and attitude data.

[0055] 2. Multi-source data fusion and control signal generation link

[0056] The data processing module 200 builds a three-stage processing pipeline to achieve intelligent conversion from raw data to control instructions:

[0057] 1. Spatiotemporal registration and noise suppression:

[0058] The data fusion unit 203 first establishes a water flow disturbance model 2031, models the water flow velocity v as a first-order Gauss-Markov process (state transfer matrix Φ = e^(-Δt / τ), τ is the water flow related time constant), and describes the impact of random disturbances on ultrasonic ranging and water pressure sensing.

[0059] The state space model 2032 defines the state vector Where b1 and b2 are sensor bias terms. In the observation equation Zk = Hxk + Vk, the H matrix dynamically switches based on the sensor type (ultrasonic ranging corresponds to distance observation, water pressure sensor corresponds to slope observation), and Vk is zero-mean Gaussian white noise (the covariance matrix R is updated using real-time noise statistics).

[0060] By iteratively updating the state estimation through the extended Kalman filter, time synchronization and error correction of multi-source data are achieved, and the fused three-dimensional terrain coordinates (X, Y, Z, θ) are output, where θ incorporates the heading angle parameters of the nine-axis inertial measurement unit.

[0061] 2. Terrain-vibration mapping engine:

[0062] The pattern generation unit 204 queries the terrain-vibration mapping database 205, which pre-stores 128 vibration parameter combinations corresponding to typical terrains. For example:

[0063] The obstacle 3 meters in front (azimuth angle α = 0°, distance d = 3m, slope θ = 5°) is mapped to the chest vibrator 302 outputting a frequency of 100 Hz and a speed of 1.5 m / s. 2 Vibration signal with intensity and phase difference of 0;

[0064] The steep slope 5 meters behind the right leg (α = 225°, d = 5m, θ = 30°) is mapped to the vibration array 304 on the back of the right leg outputting a 50Hz frequency and 3m / s 2 The vibration signal with intensity and phase difference π is transmitted to the tactile feedback module 300 via the SPI bus.

[0065] 3. Space-Frequency-Phase 3D Coding System for Tactile Feedback

[0066] The tactile feedback module 300 builds a multi-dimensional perception system of spatial partitioning, frequency modulation, and phase encoding:

[0067] 1. Physical layout and space partitioning:

[0068] The chest vibrator 302 (8 vibration motors arranged in a ring) covers the 0-180° area in front, with each motor corresponding to a 22.5° azimuth interval;

[0069] The waist vibrator 303 (4 motors on each side) covers the left / right 45-135° area, with a single motor corresponding to a 22.5° azimuth angle;

[0070] The limb vibration array (304) is divided into 8 quadrants:

[0071] The front / back left / right part of the arm, the front / back left / right part of the leg, each quadrant corresponds to a 45° azimuth interval (such as 225° in the lower left is responded by the left vibrator on the front of the left leg).

[0072] 2. Position encoding and dynamic feedback:

[0073] Phase difference modulation:

[0074] The phase difference of each vibration device is controlled via a distributed I2C bus. A phase difference of 0 corresponds to the front (α = 0°), a phase difference of π corresponds to the rear (α = 180°), and a phase difference of π / 2 corresponds to the left / right (α = 90° / 270°). For example, when an obstacle is 45° to the right front, the phase difference between the right chest vibrator and the right waist vibrator is set to π / 4, forming an azimuth coding gradient.

[0075] Frequency-Intensity Synergy:

[0076] The distance d is positively correlated with the vibration frequency (d = 1m → 200Hz, d = 5m → 20Hz), and the slope θ is positively correlated with the vibration intensity (θ = 10° → 1m / s 2 , θ=45°→5m / s 2 ), information decoupling is achieved through multi-parameter combination.

[0077] 4. Polarization-underwater acoustic fusion technology for multi-channel information transmission

[0078] The Multi-Channel Information Transmission Module 400 breaks through the bandwidth limitations of traditional underwater acoustic communications and enables efficient transmission of high-dimensional data:

[0079] 1. Polarization imaging coding layer:

[0080] Stokes parameter mapping:

[0081] The obstacle distance d is mapped to the polarization degree P (P = S1 / √(S0 2 +S1 2 +S2 2 ), range 0-1), d=0→P=1, d=10m→P=0.1;

[0082] The terrain slope θ is mapped to the polarization angle ψ (ψ = 0.5 arctan (S2 / S1)), θ = 0° → ψ = 0°, θ = 90° → ψ = 45°;

[0083] The attitude data (heading angle ψ, pitch angle θ, roll angle φ) are converted into Stokes vector components (S1, S2, S3) through the rotation matrix transformation (R = Rz(ψ)Ry(θ)Rx(φ)), where S3 corresponds to the 0° polarization state.

[0084] QPSK modulation:

[0085] The Stokes parameters are mapped to four polarization states of 45°, 135°, 90°, and 0° (corresponding to the four symbols of QPSK). Each symbol carries 2 bits of information, and the spectrum efficiency is improved to 2bps / Hz.

[0086] 2.OFDM underwater acoustic transmission layer:

[0087] Signal modulation process:

[0088] The fused polarization composite signal is converted into an optical signal by the electro-optical conversion module 403, the polarization information is loaded by the liquid crystal polarization modulator 401, and then modulated into an underwater acoustic signal by the OFDM underwater acoustic transmitter 402 (number of subcarriers 256, bandwidth 10kHz), and transmitted to the external terminal 500 through the transducer 404 (operating frequency 100-500kHz).

[0089] 3. 3D model building engine:

[0090] After receiving the signal, the external terminal 500 recovers the multimodal data using a polarization demodulation algorithm. This data is then discretized into three-dimensional space using a voxel grid algorithm (voxel unit size 0.1m×0.1m×0.1m, with an occupancy threshold of 0.6). When an obstacle occupies more than 60% of the volume within a voxel unit, it is marked as "occupied," ultimately generating a high-precision, real-time 3D terrain model.

[0091] Example 2:

[0092] In view of the above embodiment 1, further description is given in conjunction with the deep sea cliff avoidance navigation scenario, see Figures 1-6 :

[0093] Scenario Challenge

[0094] In the seamount area (water depth of 300-800 meters), there are steep cliffs with a slope of >45°. Traditional sonar can only detect horizontal distance and is not very convenient for sensing vertical drop. Submersibles can easily fall into cliffs due to misjudgment of the terrain.

[0095] Multimodal perception collaborative process

[0096] 1. 3D environment data collection

[0097] Phased array ultrasonic ranging unit 101: The 16-channel transducer array scans with a ±60° fan-shaped beam. Through the delay-add beamforming algorithm, it completes sampling of the 10-meter fan-shaped area in front within 50ms, and measures the horizontal distance d = 6 meters to the cliff edge (calculated by time-of-flight method); the echo intensity distribution shows a sudden change in sound pressure in the vertical direction (attenuation > 20dB), which is preliminarily judged to be an area with abrupt terrain changes.

[0098] Micro water pressure sensor array 102: 8 groups of MEMS sensors (10mm spacing matrix) detect the pressure difference ΔP = 0.08kPa between adjacent sensors. The terrain gradient G = ΔP / (ρgh) = 0.8rad / m (ρ is the seawater density, h is the sensor spacing) is calculated through the differential pressure algorithm, and the corresponding slope θ = arctan(G*10mm) = 38° (close to the cliff critical value).

[0099] The nine-axis inertial measurement unit 103 outputs the submersible's pitch angle φ=12° (dive tendency) and roll angle ψ=8° (tilt to the right), and combines the water pressure data to determine that the submersible is diving towards the cliff.

[0100] 2. Data fusion and risk assessment

[0101] Adaptive Kalman filter process: The water disturbance model 2031 inputs the current flow velocity u = 1.2 m / s (first-order Gauss-Markov process modeling), and the state space model 2032 updates the state vector as:

[0102] x k =[X=50m,Y=30m,Z=-500m,θ=38,b x =0.05,b y =0.03] T (where b x ,b y is the sensor bias term); the observation equation Z k =Hx k +V k In the H matrix, distance, slope, and attitude components are extracted. After extended Kalman filtering, the output fusion slope θ = 41°

[0103] (38° before correction), confirming that the cliff risk threshold has been reached (preset ≥40°).

[0104] 3. Analysis of tactile feedback spatial encoding

[0105] Obstacle orientation calculation

[0106] Ultrasonic echo peak azimuth angle α = 330° (right rear), combined with the tactile module spatial encoding rules:

[0107] The right side vibrator 303 of the waist corresponds to 45-135° to the right, but α = 330° belongs to the posterior quadrant of the limbs;

[0108] The vibration array 304 on the back of the right leg is triggered, with a phase difference of φ = 3π / 4 (corresponding to 270° + 45° = 315°, and a 15° error from 330° is within the allowable range);

[0109] Multi-parameter composite feedback:

[0110] Frequency f = 180 Hz (corresponding to a distance of 6 meters, close to the threshold of 5 meters), intensity a = 4 m / s 2 (corresponding to a slope of 41°, high-risk level), phase difference φ = 3π / 4 (right rear orientation);

[0111] The chest vibrator synchronizes low-frequency vibration (20Hz) as a global warning, and vibrations in the limbs locate the specific direction.

[0112] 4. Comparison with traditional solutions

[0113]

[0114] Example 3:

[0115] In view of the above embodiment 1, further description is given in conjunction with the underwater cave multi-target obstacle avoidance scenario, see Figures 1-6 :

[0116] Scenario Challenge

[0117] In underwater cave networks (50-100 meters deep, with significant multipath effects), it is necessary to simultaneously avoid top stalactites (2 meters away), side rock walls (60° slope) and bottom sedimentary layers (soft terrain requires controlling the descent speed). Traditional underwater acoustic communications are not convenient for real-time transmission of three-dimensional data, resulting in remote modeling delays of >10 seconds.

[0118] The entire process of multi-channel information transmission

[0119] 1. Multimodal Data Encoding Rules

[0120] Polarization state-parameter mapping:

[0121] Stalactite distance d = 2m → polarization degree P = S1 / (S0) = 0.6 (corresponding to a 45° polarization state);

[0122] Rock face slope θ = 60° → polarization angle χ = arctan(S2 / S1) = 30° (corresponding to a 135° polarization state, since S2 = tan(θ) × S1);

[0123] Submersible attitude (pitch angle φ = 5°) → Stokes parameter S3 = sin(2φ) = 0.17 (corresponding to the 90° polarization component);

[0124] OFDM modulation parameters: 512 subcarriers, QPSK modulation order, 10ms symbol period, and cyclic prefix (CP=2ms) to suppress inter-symbol interference in a multipath environment.

[0125] 2. Communication link anti-interference mechanism

[0126] Underwater acoustic channel equalization: After the transducer 404 receives the signal, it performs channel estimation through the UWM1000 protocol stack, uses the least squares method to fit the multipath delay (τ1 = 5ms, τ2 = 8ms), corrects the polarization parameter phase offset, and the bit error rate is reduced from the initial 10 -3 Down to 10 -5 .

[0127] 3. Real-time 3D terrain modeling

[0128] Voxel grid algorithm implementation:

[0129] After receiving the polarized underwater acoustic signal, the external terminal 500 performs modeling according to the following steps:

[0130] 1. Polarization demodulation to obtain S1 / S2 / S3 parameters and inversely calculate distance, slope, and attitude;

[0131] 2. The three-dimensional space is discretized into 0.5m×0.5m×0.5m voxel units;

[0132] 3. Marking voxel occupancy for stalactite areas = 0.9 (solid obstacles), rock wall areas = 0.7 (steep terrain), and sedimentary layers = 0.3 (passable but with caution). Modeling delay is reduced from 12 seconds in the traditional solution to 1.8 seconds.

[0133] 4. Multi-target collaborative obstacle avoidance strategy

[0134] Haptic feedback synchronization update:

[0135] Top stalactite → chest vibrator high frequency (200Hz), high intensity (5m / s 2 )shock;

[0136] Side rock wall → left arm rear vibration array phase difference π (corresponding to 180°±45° to the left rear, need to turn right);

[0137] Bottom sediment layer → Low frequency (30Hz) vibration array on the front of the leg prompts "slow rise".

[0138] Example 4:

[0139] In view of the above embodiment 1, further description is given in combination with the detailed mapping scenario of underwater archaeological sites, see Figures 1-6 :

[0140] Scenario Challenge

[0141] At ancient shipwreck sites (water depth 40 meters, current velocity 0.5m / s), it is necessary to map the wreckage (30 meters long, 5 meters high), scattered artifacts (diameter <0.3 meters), and the surrounding sandy microtopography (slope <5°). Traditional systems lack resolution, resulting in a >30% artifact miss rate.

[0142] Technology chain collaboration

[0143] 1. High-precision perception mode

[0144] Phased array ultrasound parameter adjustment:

[0145] Switching to high-resolution mode: beam width compressed to 2° (conventional ±60° sector scan changed to narrowband scan), sampling interval 0.1m, a 0.2m diameter pottery jar was detected at the hull fracture (range accuracy ±5cm, time-of-flight + beamforming analysis);

[0146] Micro-topography capture by the water pressure sensor array: 8 groups of sensors detected a pressure difference of 0.01 kPa (corresponding to a height difference of 0.02 meters), and calculated that there was a 0.1-meter-high protrusion on the sand (possibly a fragment of a cultural relic).

[0147] 2. Data fusion and deep processing

[0148] Adaptive Kalman filter iteration: The water disturbance model is updated to a second-order Gauss-Markov process, and the flow velocity gradient is estimated in real time. The state-space model incorporates the size characteristics of cultural relics (radius r = 0.1-0.5m) and realizes small target trajectory tracking through extended Kalman filtering.

[0149] 3. Haptic feedback graded warning

[0150] Cultural Relics-Topography Classification Code:

[0151] Ceramic pot (small target) → Single-point trigger of the limb vibration array (e.g., the front of the right hand corresponds to 90 degrees to the right front), frequency 150Hz (medium risk), intensity 1.5m / s 2 ;

[0152] Hull wreckage (large obstacle) → Chest + waist vibrator linkage, phase difference 0 (directly in front), frequency 50Hz (long-range warning);

[0153] Micro-topography tips:

[0154] Sand bump → front leg vibration array low frequency (20Hz), low intensity (0.5m / s 2 ), prompting "Slight terrain change, move with caution."

[0155] 4. Comparison of surveying and mapping results

[0156] index Traditional sonar + vision combination The system and method Minimum detection size 0.5m (missing inspection of pottery jars) 0.1m (complete capture of 0.2m pottery jar) Microtopography resolution 0.5m height difference 0.02m height difference Modeling time 45 minutes (manual splicing) 8 minutes (real-time voxel mesh construction)

[0157] The above description is merely a preferred embodiment of the present invention; however, the scope of protection of the present invention is not limited thereto. Any person skilled in the art who, within the technical scope disclosed by the present invention, makes equivalent substitutions or modifications based on the technical solutions and improved concepts of the present invention shall be covered by the scope of protection of the present invention.

Claims

1. A multimodal underwater terrain information perception system, characterized by: It comprises a multimodal perception module (100), a data processing module (200), a tactile feedback module (300) and a multi-channel information transmission module (400); The multimodal sensing module (100) is connected to the input port of the data processing module (200) via an SPI bus data communication connection. The multimodal sensing module (100) comprises a phased array ultrasonic ranging unit (101), a micro water pressure sensor array (102), and a nine-axis inertial measurement unit (103), and is used to collect multimodal data of obstacle distance, terrain slope, and diver's posture. The data processing module (200) comprises an embedded processor (201) and a storage unit (202). The data processing module (200) is connected to the input port of the tactile feedback module (300) via an SPI bus for data communication. The tactile feedback module (300) interacts with the multi-channel information transmission module (400) in real time via a UWM1000 underwater acoustic communication protocol stack to transmit execution status data of the vibration control signal. The data processing module (200) further comprises a data fusion unit (203), a pattern generation unit (204) and a pre-stored terrain-vibration mapping database (205). The data fusion unit (203) uses an adaptive Kalman filter algorithm to perform spatiotemporal calibration on the received multimodal data. The pattern generation unit (204) generates a vibration control signal comprising quantitative parameters of obstacle azimuth α, distance d and terrain slope θ according to the pre-stored terrain-vibration mapping database (205). The signal parameters include a frequency of 20-200 Hz, a speed of 0.1-5 m / s, and a speed of 10-200 Hz. 2 intensity, 0-π phase difference; The tactile feedback module (300) is connected to a vibration device (301) via a distributed I2C bus. The vibration device (301) comprises a chest vibrator (302), a waist vibrator (303), and a limb vibration array (304). Each of the vibration devices (301) has a built-in micro-vibration motor and a drive circuit. According to the control signal output by the data processing module (200), spatial coding prompts of obstacle positions are achieved through the spatial distribution and vibration phase difference of the vibration devices (301) in different areas. The multi-channel information transmission module (400) comprises a liquid crystal polarization modulator (401), an OFDM underwater acoustic transmitter (402), an electro-optical conversion module (403) and a transducer (404); the transducer (404) is connected to an external terminal (500) via an SPI bus; the electro-optical conversion module (403) encodes processed multimodal data into an optical signal containing polarization degree and polarization angle information, converts the optical signal into an underwater acoustic signal via the transducer (404) and transmits the signal to the external terminal (500); and the external terminal (500) constructs a three-dimensional underwater terrain model based on the received signal using a voxel grid algorithm.

2. The underwater terrain information multimodal perception system according to claim 1, characterized in that: The phased array ultrasonic ranging unit (101) comprises a 16-channel ultrasonic transducer array (1011), adopts a delay-add beamforming algorithm to form a ±60° fan-shaped detection area, and cooperates with the time-of-flight method to calculate obstacle distance data. The delay-add beamforming algorithm realizes beam direction control through the delay time difference of each channel ultrasonic transducer. The micro water pressure sensor array (102) comprises 8 groups of MEMS piezoresistive sensors (1021), and the 8 groups of MEMS piezoresistive sensors (1021) are distributed on the detection surface in a two-dimensional matrix with a spacing of 10 mm. The terrain slope is solved by a differential pressure algorithm, and the differential pressure algorithm calculates the terrain gradient through the pressure difference of adjacent sensors.

3. The underwater terrain information multimodal perception system according to claim 1, characterized in that: The data fusion unit (203) establishes a water flow disturbance model (2031) and a state space model (2032). The water flow disturbance model (2031) is a first-order Gauss-Markov process that describes the random interference of water flow velocity on sensor measurement. The state vector of the state space model (2032) includes three-dimensional coordinates, slope θ and sensor deviation terms. The observation equation of the state space model (2032) is: z k =Hx k +v k , Where H is the observation matrix, v k The method is a method for realizing time synchronization and error correction of multi-source data by using an extended Kalman filter, and outputting fused three-dimensional terrain coordinate data in a format of (X, Y, Z, θ), wherein θ represents a heading angle parameter calculated based on the nine-axis inertial measurement unit (103).

4. The underwater terrain information multimodal perception system according to claim 1, characterized in that: The vibration pattern of the tactile feedback module (300) includes a spatial coding rule, wherein the chest vibrator (302) corresponds to the 0-180° area in the front, the waist vibrator (303) corresponds to the 45-135° area on the left / right, and the limb vibration array (304) is distributed in four quadrants, namely, the front of the arm, the back of the arm, the front of the leg, and the back of the leg, wherein the left front of the arm corresponds to 90° to the left, the right front of the arm corresponds to 270° to the right, the left front of the leg corresponds to 225° to the lower left, and the right front of the leg corresponds to 315° to the lower right. The obstacle orientation information is transmitted through the phase difference 0-π of the vibrators in different quadrants. When the phase difference is 0, it corresponds to 0° in the front, when the phase difference is π, it corresponds to 180° on the back, and when the phase difference is π / 2, it corresponds to 90° to the left or 270° to the right.

5. The underwater terrain information multimodal perception system according to claim 1, characterized in that: The multi-channel information transmission module (400) adopts polarization imaging and underwater acoustic communication fusion technology, supports QPSK modulation, and modulates terrain data into four polarization states of 45°, 135°, 90°, and 0° through Stokes parameters, wherein Stokes parameter S1 corresponds to the 45° polarization state, i.e., obstacle distance, S2 corresponds to the 135° polarization state, i.e., terrain slope, S0 corresponds to the 90° polarization state, and S3 corresponds to the 0° polarization state, i.e., attitude data component, and underwater data transmission is realized by using orthogonal frequency division multiplexing technology.

6. A multimodal perception method for underwater terrain information, comprising a multimodal perception system for underwater terrain information according to any one of claims 1 to 5, characterized in that: The following steps are involved: S1. Data acquisition and synchronization: The multimodal sensing module (100) is based on the PTP precision time protocol and is connected to the PTP modules of each sensor slave node through a PTP master node integrated in the embedded processor (201) and an independent synchronization control line. Based on the IEEE 1588 protocol, the phased array ultrasonic ranging unit (101), the micro water pressure sensor array (102) and the nine-axis inertial measurement unit (103) are synchronized at the nanosecond level, and the distance, slope and attitude data are synchronously collected and transmitted to the data processing module (200) via the SPI bus; S2, data fusion and signal generation: The data processing module (200) performs spatiotemporal registration on the multi-source data through the data fusion unit (203), eliminates water flow noise and sensor drift using the adaptive Kalman filter algorithm, matches the terrain-vibration mapping database (205) through the pattern generation unit (204), and generates a vibration control signal containing spatial positioning information; S3. Information transmission: The tactile feedback module (300) drives the vibration device (301) at the corresponding part according to the control signal, and transmits obstacle distance, direction and terrain slope information to the user through a combination of differentiated vibration frequency, intensity and phase difference; S4, polarization imaging coding: The multi-channel information transmission module (400) encodes the fused terrain data into a polarization image signal through a liquid crystal polarization modulator (401), modulates the signal through an OFDM underwater acoustic transmitter (402), and transmits the signal to an external terminal (500) through a transducer (404). The external terminal (500) constructs a three-dimensional terrain model based on polarization demodulation and a voxel grid algorithm. The voxel grid algorithm is based on three-dimensional space discretization and characterizes terrain features through the occupancy rate of voxel units.

7. The multimodal underwater terrain information perception method according to claim 6, characterized in that: The polarization imaging encoding process described in step S4 is as follows: by mapping the obstacle distance to the polarization degree, mapping the terrain slope to the polarization angle, and mapping the attitude data to the Stokes vector component through the rotation matrix transformation from the Euler angle to the Cartesian coordinate system, a polarization composite signal containing multimodal information is formed, which is loaded onto the OFDM subcarrier for transmission after QPSK modulation. The rotation matrix transformation formula is: R(α, β, γ) = R z (γ)R y (β)R x (α).

8. The multimodal perception method for underwater terrain information according to claim 6, characterized in that: The voxel grid algorithm in step S4 is based on three-dimensional space discretization, the voxel unit size is Δx×Δy×Δz=0.1m×0.1m×0.1m, the occupancy threshold is set to 0.6, and the terrain characteristics are characterized by the occupancy of the voxel unit.

Images