Inland river unmanned ship navigation anti-collision method and system based on multi-point ultrasonic distance measurement

Abstract

The invention relates to the technical field of inland river shipping, in particular to an inland river unmanned ship navigation anti-collision method and system based on multipoint ultrasonic ranging, and the method comprises the steps: configuring an ultrasonic sensor array which comprises five ultrasonic sensor nodes; original ranging signals of all the ultrasonic sensor nodes are circularly collected based on a time division multiplexing mechanism; constructing a virtual radar matrix; executing dynamic threat evaluation; and performing obstacle avoidance execution control. Five ultrasonic sensor nodes are configured to form a 180-degree fan-shaped array, and a continuous sensing field covering a bow and two side wings is formed in cooperation with a time division multiplexing mechanism and a virtual radar matrix algorithm; wherein co-frequency cross interference between ultrasonic sensors is eliminated through a time division multiplexing strategy, sensing discontinuity between discrete detection points is eliminated through a spatial interpolation technology, and the problem of lateral sensing blind areas existing in the background technology is solved.

Description

Technical Field

[0001] This application relates to the field of inland waterway shipping technology, and in particular to a collision avoidance method and system for unmanned inland waterway vessels based on multi-point ultrasonic ranging. Background Technology

[0002] With the intelligent development of shipping technology, unmanned inland waterway vessels are increasingly widely used in environmental inspection, material transportation and underwater surveying. Inland waterways are usually characterized by narrow channels, variable water flow speeds and high traffic density, which poses a severe challenge to the autonomous perception and real-time obstacle avoidance capabilities of unmanned vessels in complex environments.

[0003] In the prior art, a fusion obstacle avoidance method combining millimeter-wave radar and ultrasonic sensors is disclosed. The target information is obtained by millimeter-wave radar, and the ultrasonic sensor is used for auxiliary detection within a detection angle range of ±30 degrees. The obstacle avoidance command is executed by judging whether the distance and angle between the obstacle and the hull meet the preset threshold. Another prior art discloses a collision avoidance algorithm optimized for inland waterway environments. It establishes a mathematical model of obstacles by improving the speed obstacle method and makes collision avoidance decisions based on the speed vector conflict area.

[0004] However, the aforementioned existing technologies still have limitations in practical applications. When dealing with narrow inland waterways, due to the layout logic of the sensors, there are obvious blind spots in the lateral sensing area, making it difficult to obtain the real-time distance between the side of the hull and the shoreline or obstacles. The simultaneous operation of multiple ultrasonic sensors in a limited space will generate severe cross-interference of the same frequency, leading to false alarms in the sensing data. Existing solutions lack the means to convert discrete ranging data into a continuous spatial distribution field, making it difficult to complete high-precision obstacle avoidance decisions under a low-cost sensing framework. Furthermore, there is still room for improvement in the dynamic threat quantification assessment and deep coupling control of the propulsion system. Summary of the Invention

[0005] The objective of this application is to provide a collision avoidance method for unmanned inland waterway vessels based on multi-point ultrasonic ranging, comprising: configuring an ultrasonic sensor array, wherein the ultrasonic sensor array includes five ultrasonic sensor nodes, the five ultrasonic sensor nodes being respectively installed in a fan-shaped topology at the zero-degree central axis position, at a positive 45-degree oblique position and a negative 45-degree oblique position symmetrically distributed relative to the central axis, and at a positive 90-degree lateral position and a negative 90-degree lateral position symmetrically distributed relative to the central axis; cyclically acquiring the original ranging signals of each ultrasonic sensor node based on a time-division multiplexing mechanism; constructing a virtual radar matrix, normalizing and mapping the discrete original ranging signals and projecting them onto a two-dimensional polar coordinate space. In the inter-model, virtual ranging values ​​for each detection gap point are calculated using a spatial interpolation algorithm to generate a continuous obstacle detection field distribution map. Dynamic threat assessment is performed, extracting the obstacle center distance, relative approach rate, and azimuth deviation relative to the ship's heading from the obstacle detection field distribution map. A quantified comprehensive threat assessment coefficient value is obtained through a nonlinear weighted formula. Obstacle avoidance execution control is implemented, triggering obstacle avoidance actions in response to the comprehensive threat assessment coefficient value. A cascaded proportional-integral-derivative controller converts the comprehensive threat assessment coefficient value into corresponding track deviation correction and propulsion power adjustment amounts. The propulsion power adjustment amounts include forward speed adjustment commands for the unmanned vessel's propulsion system and reverse braking commands in emergency obstacle avoidance situations.

[0006] By adopting the above technical solution, five ultrasonic sensor nodes are configured to form a 180-degree sector array. Combined with the time-division multiplexing mechanism and the virtual radar matrix algorithm, a continuous sensing field covering the bow and both sides of the ship is formed. Among them, the time-division multiplexing strategy eliminates the cross-interference between ultrasonic sensors at the same frequency, and the spatial interpolation technology eliminates the discontinuity of perception between discrete detection points, thus solving the problem of lateral sensing blind zone in the background technology.

[0007] Optionally, the process of constructing the virtual radar matrix includes: establishing a spatial grid model in a polar coordinate system, wherein the angle step size of the spatial grid model is configured to be 1 to 5 degrees; marking the original ranging signal on the corresponding zero-degree, ±45-degree, and ±90-degree radial axes; calculating the depth values ​​of empty grid cells between adjacent radial axes using a bilinear interpolation algorithm; generating a continuous ranging envelope covering a 180-degree range in front of the unmanned vessel to form a two-dimensional heat map reflecting the spatial morphology of obstacles.

[0008] By adopting the above technical solution and introducing a bilinear interpolation algorithm into the polar coordinate space model, one-dimensional discrete pulse data is transformed into a two-dimensional heat map with spatial breadth, resulting in the technical effect of acquiring high-dimensional sensing information similar to millimeter-wave radar using low-cost ultrasonic hardware.

[0009] Optionally, the method further includes a dynamic clutter filtering process, which includes: extracting time-domain features from the two-dimensional heat map; identifying and filtering transient noise with a duration less than a preset time threshold in the ranging data based on the frequency characteristics of the inland river wave motion; and using a median filtering algorithm to process high-frequency artifacts in the detection field distribution map to eliminate false triggering signals caused by water splashing.

[0010] By adopting the above technical solution and using wave frequency characteristics for time-domain feature extraction, transient noise caused by water splashes and clutter is effectively filtered out, improving the system's false alarm resistance in harsh water environments.

[0011] Optionally, the process of cyclic acquisition based on time-division multiplexing mechanism includes: configuring a clock synchronization controller to divide the five ultrasonic sensor nodes into non-interfering transmission groups; cyclically sending ultrasonic pulse trigger commands in the order of zero-degree center position, ±45-degree position, and ±90-degree position; and setting a preset time delay interval between the trigger commands of adjacent groups, wherein the preset time delay interval is configured to be greater than twice the round-trip propagation time of the ultrasonic wave within a preset maximum range.

[0012] By adopting the above technical solution and setting a preset time delay interval greater than twice the maximum range propagation time, the physical isolation of adjacent sound wave pulses on the time axis is ensured, eliminating the risk of sound wave reflection overlap in narrow spaces.

[0013] Optionally, the calculation process of the comprehensive threat assessment coefficient includes: extracting the reciprocal of the obstacle distance as a first weighting factor; extracting the relative approach velocity as a second weighting factor; extracting the obstacle azimuth deviation as a third weighting factor; and performing a nonlinear summation operation on the first weighting factor, the second weighting factor, and the third weighting factor; wherein the comprehensive threat assessment coefficient is positively correlated with the degree of collision risk.

[0014] By adopting the above technical solution, a quantitative collision probability model was established by nonlinearly weighting and integrating distance, speed, and orientation dimensions, thus enabling obstacle avoidance decisions to have dynamic predictive attributes.

[0015] Optionally, it also includes a lateral distance balance control process, specifically including: extracting the ranging deviation of the ultrasonic sensor node at a positive 90-degree lateral position and a negative 90-degree lateral position in real time; calculating the difference in distance to obstacles on both sides of the hull in response to the unmanned vessel being in berthing mode or narrow waterway passage mode; and inputting the difference into a lateral drift compensation algorithm to generate a lateral dynamic compensation command for adjusting the berthing attitude.

[0016] By adopting the above technical solution, the lateral distance maintenance accuracy of unmanned vessels is improved under restricted waterway conditions such as berthing and crossing gates by extracting the distance deviation of positive and negative 90 degree nodes and applying it to lateral drift compensation.

[0017] Optionally, the implementation process of the cascaded proportional-integral-derivative controller includes: establishing a closed-loop control structure with the heading angle deviation as the outer loop and the thruster speed as the inner loop; dynamically adjusting the proportional coefficient, integral coefficient and derivative coefficient of the controller according to the magnitude of the comprehensive threat assessment coefficient; and prioritizing the correction of the unmanned vessel's heading by adjusting the rudder angle under the condition of meeting the preset safe distance.

[0018] Optionally, the process also includes a heading smoothing process, specifically including: configuring a low-pass filter at the output of the cascaded proportional-integral-derivative controller; setting the dead zone range for heading correction, and maintaining the current thruster state unchanged when the track deviation correction amount is less than the preset minimum dead zone value; and eliminating heading oscillations generated during continuous obstacle avoidance by limiting the maximum slope of the rudder angle change rate.

[0019] By adopting the above technical solution and limiting the slope of the rudder angle change rate, the control oscillation caused by sudden changes in obstacle avoidance commands is eliminated, ensuring the smooth navigation of the unmanned vessel under continuous obstacle avoidance conditions.

[0020] Optionally, it also includes an environmental baseline compensation process, specifically including: real-time acquisition of atmospheric temperature and humidity data; calculation of real-time sound speed values ​​under the current atmospheric environment based on the sound speed propagation correction formula; and using the real-time sound speed values ​​to perform distance compensation correction on the original ranging signal to eliminate the spatial positioning deviation of the virtual radar matrix caused by changes in ambient temperature and humidity.

[0021] By adopting the above technical solution and using real-time temperature and humidity compensation for the sound speed reference, the absolute accuracy of ranging of the virtual radar matrix in different seasons and climates is guaranteed.

[0022] The second objective of this application is to provide a collision avoidance system for inland waterway unmanned vessels based on multi-point ultrasonic ranging, comprising: a memory; a processor; and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the computer program, implements the collision avoidance method for inland waterway unmanned vessels based on multi-point ultrasonic ranging as described above. Attached Figure Description

[0023] Figure 1 This is a schematic diagram of the collision avoidance method for inland waterway unmanned vessels based on multi-point ultrasonic ranging, as described in this application. Detailed Implementation

[0024] The following will provide a complete description of the collision avoidance method and system for inland waterway unmanned vessels based on multi-point ultrasonic ranging, using specific embodiments of this application, so that those skilled in the art can fully understand and implement all the technical contents of this application. This application is primarily aimed at the specific application scenario of inland waterways, which typically features narrow waterways, high traffic density, and various obstacles such as fixed bridge piers and floating debris. Traditional obstacle avoidance schemes based on a single forward sensor are prone to scraping or collisions when dealing with nearby embankments or other vessels suddenly appearing on the side of the vessel due to the existence of lateral perception blind spots. Furthermore, when multiple ultrasonic sensors are densely deployed in the limited bow space, if a synchronous triggering mode is used, the 40 kHz ultrasonic pulses emitted by the sensors will interfere with each other in the air and form [missing information]. Complex reflection superposition makes it impossible for the receiver to accurately resolve the actual obstacle distance corresponding to the echo signal, resulting in serious co-channel cross-interference problems and causing a large number of false alarms in the sensing data. In addition, discrete ranging points can only provide linear distance information in various directions, and cannot form a continuous description of the spatial shape such as the outline and width of the obstacle. This makes it difficult for the unmanned vessel's control system to make fine obstacle avoidance decisions based on incomplete spatial information. For example, it cannot determine whether it is safer to go around the obstacle from the left or right side, nor can it accurately maintain equidistant navigation between the hull and the gate walls on both sides when passing through narrow gates.

[0025] like Figure 1 As shown, this application provides a collision avoidance method for unmanned inland waterway vessels based on multi-point ultrasonic ranging, including the following steps.

[0026] S01: Configure an ultrasonic sensor array, which includes five ultrasonic sensor nodes. The five ultrasonic sensor nodes are installed in a fan-shaped topology at the zero-degree central axis position, at a positive 45-degree oblique position and a negative 45-degree oblique position symmetrically distributed with respect to the central axis, and at a positive 90-degree lateral position and a negative 90-degree lateral position symmetrically distributed with respect to the central axis.

[0027] Specifically, the hull of the unmanned vessel can be made of fiberglass, for example, with a total length of 1.2 meters, a width of 0.4 meters, and a draft of 0.15 meters. This size design enables it to meet the passage requirements of most inland tributaries and narrow waterways.A sensor mounting bracket is fixedly installed 0.2 meters above the bow deck of the unmanned vessel. This bracket is made of aluminum alloy, which has excellent rigidity and corrosion resistance. Five ultrasonic sensor nodes are mounted on the bracket in a fan-shaped topology. The specific model of the ultrasonic sensor node is HC-SR04, with an operating frequency of 40 kHz, a maximum theoretical range of 400 cm, a minimum detectable distance of 2 cm, and an operating voltage of 5 V DC. The installation positions of the five ultrasonic sensor nodes are spatially calibrated: the first node is installed at the zero-degree center axis of the bow of the unmanned vessel, with its acoustic emission surface strictly perpendicular to the longitudinal axis of the hull and pointing in the direction of travel; the second and third nodes are symmetrically distributed relative to the center axis, installed at a 45-degree and a -45-degree angle respectively, with the installation angle error controlled within ±1 degree to ensure the symmetry of the oblique detection area. The fourth and fifth nodes are also symmetrically distributed relative to the central axis, installed at positive 90-degree and negative 90-degree lateral positions respectively, with their acoustic emission surfaces strictly pointing to the port and starboard sides of the hull. The fan-shaped topology covers a 180-degree horizontal space in front of the unmanned vessel. The zero-degree node is responsible for detecting long-range obstacles directly ahead, the ±45-degree nodes are responsible for detecting mid-range obstacles on the port and starboard sides, and the ±90-degree nodes are specifically used to detect shorelines, docks, or other vessels close to the sides of the hull. This achieves physical coverage of the bow and flank danger zones at the hardware level. The ultrasonic sensor nodes are connected to the ship's main control unit via a four-core shielded cable. Two lines provide 5V DC power, and the other two lines are used to receive trigger signals and send echo signals, respectively. The cable shield is grounded to suppress electrical noise interference. The core processor of the ship's main control unit can be based on ARM. The microcontroller, powered by a Cortex-M4 core with a clock speed of 120 MHz, incorporates an analog-to-digital converter and multiple general-purpose timer units, enabling it to meet the demands of high-speed signal processing and precise timing control. The propulsion system comprises two brushless DC motors, each driving a propeller on the port and starboard sides via independent drive shafts. These motors are controlled by a dual-channel DC motor drive module (model TB6612), which can receive pulse-width modulation (PWM) signals to precisely adjust motor speed and direction. The steering mechanism utilizes a metal gear servo (model MG996R) with an output torque of 10 kgf / cm, driving the rudder blade connected to the stern. The rudder blade has a maximum deflection angle of ±30 degrees. The entire system is powered by a 12-volt, 10-amp-hour lithium polymer battery pack, which generates 5-volt and 3.3-volt voltage rails via a DC-DC converter module to power the sensors, servos, and main control unit.

[0028] S02: Based on the time-division multiplexing mechanism, the original ranging signals of each ultrasonic sensor node are collected cyclically.

[0029] Specifically, the process of cyclic acquisition based on the time-division multiplexing mechanism may include: configuring a clock synchronization controller to divide the five ultrasonic sensor nodes into non-interfering transmission groups; cyclically sending ultrasonic pulse trigger commands in the order of zero-degree center position, ±45-degree position, and ±90-degree position; and setting a preset time delay interval between the trigger commands of adjacent groups, wherein the preset time delay interval is configured to be greater than twice the round-trip propagation time of the ultrasonic wave within the preset maximum range.

[0030] Understandably, the entire method operates on a 100-millisecond cycle, meaning it completes a full closed loop from data acquisition to control command output every 0.1 seconds. Time segmentation is used to avoid cross-interference caused by multiple ultrasonic sensors operating simultaneously. The clock synchronization controller within the main control unit is configured with a 500-millisecond acquisition cycle and divides the five ultrasonic sensor nodes into three non-interfering transmission groups. The first group contains only the sensor node located at the zero-degree center axis position; the second group contains two sensor nodes located at a +45° and -45° oblique positions; and the third group contains two sensor nodes located at a +90° and -90° lateral positions. The acquisition process cyclically sends ultrasonic pulse trigger commands in the order of the zero-degree center position group, the +45° and +90° position groups. Between trigger commands from adjacent groups, the system sets a preset time delay interval, which is configured to be greater than twice the round-trip propagation time of the ultrasonic wave within a preset maximum range. Specifically, considering that the speed of sound in dry air at 20 degrees Celsius is approximately 340 meters per second, for a maximum range of 400 centimeters (4 meters), the one-way propagation time is approximately 11.76 milliseconds, and the round-trip time is approximately 23.52 milliseconds. Therefore, the preset time delay interval can be set to 50 milliseconds (one embodiment of the preset time delay interval). This value provides sufficient margin to handle potential issues with sound waves in complex inland waterway environments (such as those with bulkheads, railings, etc.). While handling complex situations such as multiple reflections and reverberation, the system also ensures real-time response, guaranteeing that data acquisition from all five nodes can be completed within 500 milliseconds. After each sensor node is triggered, the main control unit starts its internal high-precision timer with a timing resolution of 1 microsecond, waiting to receive the echo signal returned by the sensor. Once a valid rising edge of the echo signal is received, the timing is stopped immediately, and the timing value is multiplied by the real-time sound speed in the current environment and then divided by 2 to obtain the original distance value from the sensor to the obstacle. This original distance value is the original ranging signal, and its data type is a 32-bit floating-point number, with the unit being meters.

[0031] It should be noted that the aforementioned 500-millisecond complete data acquisition cycle and 100-millisecond system control cycle design are optimized results after comprehensively considering the dynamic characteristics of the inland waterway navigation environment and the real-time requirements of the system. In inland waterways, the water flow speed is typically between 0.5 meters per second and 3 meters per second, and the relative speeds of other vessels or floating obstacles are also in this range. The frequency of significant changes in the spatial position of obstacles is usually less than 10 Hz. Therefore, a 100-millisecond control cycle means that the system can complete the closed loop of perception-decision-execution within the centimeter-level changes in environmental conditions. Its response speed is much faster than the rate of change of state of typical obstacles in inland waterways, which is sufficient to ensure navigation safety. The 500-millisecond data acquisition cycle is to completely eliminate... Co-frequency interference between ultrasonic sensors provides ample time margin, ensuring the absolute reliability of the raw sensing data. The system adopts a hierarchical data processing architecture. The bottom-level data acquisition prioritizes absolute reliability, using a conservative 500-millisecond cycle to completely eliminate interference. The upper-level control decision-making operates on a 100-millisecond cycle, with the latest available data as its input. This latest available data may be data from a sensor group that has just been updated in the current cycle, or it may be a fusion result combining historical data. At the same time, the system can also be equipped with a high-priority interrupt channel, ensuring high reliability of sensing data under normal conditions while ensuring that the overall response speed of the system to external changes is much faster than the typical motion state change speed of obstacles in inland waterways.

[0032] Understandably, dividing the five sensor nodes into three groups and triggering them in a specific order, combined with a preset time delay interval of up to 50 milliseconds, ensures the physical isolation of adjacent acoustic pulses on the time axis, completely eliminating the risk of data confusion caused by overlapping acoustic reflections in the narrow bow space. This is a technological advancement that enables the reliable deployment of multiple ultrasonic sensors in a limited space.

[0033] In extreme cases, when a sensor detects an obstacle signal at extremely close range (e.g., less than 0.5 meters) within the current 100-millisecond cycle, the system can immediately trigger a high-priority interrupt. This interrupt handler has the highest execution priority and can preempt the main control loop within 1 to 5 milliseconds. After the interrupt is triggered, the system will immediately break out of the normal time-division multiplexing sequence, suspend the triggering of other sensors, and complete the threat assessment and generate control commands within a few milliseconds based on the emergency data.

[0034] S03: Construct a virtual radar matrix, normalize and map the discrete original ranging signals and project them into a two-dimensional polar coordinate space model, and calculate the virtual ranging values ​​of each detection gap point through a spatial interpolation algorithm to generate a continuous obstacle detection field distribution map.

[0035] Understandably, the process of constructing a virtual radar matrix includes: establishing a spatial grid model in polar coordinates, where the angular step size of the spatial grid model is configured to be one to five degrees; marking the original ranging signals on the corresponding zero-degree, ±45-degree, and ±90-degree radial axes; using a bilinear interpolation algorithm to calculate the depth values ​​of empty grid cells between adjacent radial axes; and generating a continuous ranging envelope covering a 180-degree range in front of the unmanned vessel, forming a two-dimensional heat map reflecting the spatial morphology of obstacles.

[0036] Specifically, the system first establishes a two-dimensional spatial mesh model in polar coordinates in memory. This two-dimensional spatial mesh model takes the center of the ship's hull as the origin and the direction of the bow as the zero-degree baseline, covering an 180-degree sector area from -90 degrees to +90 degrees. In this embodiment, the angle step size of the two-dimensional spatial mesh model is configured as 3 degrees, that is, 60 angle sectors are evenly divided within the sector area; the radial distance resolution is configured as 0.1 meters, and the maximum radial distance is 4 meters. Therefore, the entire two-dimensional spatial mesh model consists of 2400 mesh units, which are 60 angle sectors multiplied by 40 distance rings. The system performs normalization mapping on the acquired raw ranging signals. Specifically, the distance value measured by each sensor is divided by the maximum range of 4 meters to obtain a dimensionless ratio between 0 and 1. This dimensionless ratio is then multiplied by 40 and mapped to the corresponding distance ring index. Next, these mapped data points are marked on the corresponding angular sector axes in the two-dimensional spatial grid model: data from the zero-degree center axis sensor is marked on the zero-degree axis; data from the ±45° and ±45° sensors are marked on the ±45° and ±45° axes, respectively; and data from the ±90° and ±90° sensors are marked on the ±90° and ±90° axes, respectively. At this point, only the grid cells on these five axes contain actual ranging data; the grid cells corresponding to the remaining 55 angular sectors are data gaps. To fill these gaps and generate a continuous obstacle detection field distribution map, a bilinear interpolation algorithm can be used. For any empty grid cell located between two known data axes, such as a cell in the 30-degree sector, the system searches for its two adjacent known data axes, namely the 0-degree axis and the +45-degree axis, and obtains the data values ​​at the same distance loops on these two axes. Then, based on the proportional relationship between the angle of the empty grid cell (30 degrees) and the angles of the two known axes (0 degrees and +45 degrees), a linear weighted average is performed on the two data values ​​to calculate the virtual ranging value of the empty grid cell. By performing this interpolation calculation on all empty grid cells, a two-dimensional matrix covering a 180-degree range in front of the unmanned vessel is finally generated; this two-dimensional matrix is ​​the virtual radar matrix. It should be noted that spatial interpolation technology is essentially a data inference based on known points. For the edge of an obstacle located between two known detection directions, the virtual ranging value generated by interpolation may have certain uncertainties or smoothing effects. To address this limitation, the system can adopt a conservative strategy in the subsequent path planning and decision-making modules: for the data of virtual grid cells generated by interpolation, the system automatically adds a conservative distance margin when calculating the safe zone. For example, the safe distance threshold of the interpolation area is increased from 1 meter to 1.5 meters. This provides an additional safety buffer for possible interpolation errors when planning obstacle avoidance paths, ensuring that even if there are deviations in the interpolation inference, there will be no risk of collision.For intuitive display, the system can visualize the matrix as a two-dimensional heatmap, where the color depth represents distance, dark color represents nearby obstacles, and light color represents distant or no obstacles, thus forming a continuous ranging envelope map that reflects the spatial shape of obstacles.

[0037] In this embodiment, for any point to be interpolated in the polar coordinate grid, its interpolation weight in the angle dimension is determined by the inverse ratio of its angle difference with the two known data axes on the left and right. In the radial distance dimension, interpolation is performed on the same distance loop, that is, interpolation is performed only along the circumferential direction. This bilinear interpolation algorithm effectively transforms one-dimensional discrete pulse data into a two-dimensional continuous distribution field with spatial breadth under limited computing resources, resulting in the technical effect of acquiring high-dimensional sensing information similar to millimeter-wave radar using low-cost ultrasonic hardware.

[0038] Understandably, if the number of sensors is increased without time-division multiplexing, the co-channel interference problem will worsen dramatically with the increase in the number of sensors. Experimental data shows that when five ultrasonic sensors are simultaneously triggered in a 0.5 square meter area at the bow, the false alarm rate due to cross-interference is as high as 30%. However, after adopting the time-division multiplexing strategy of this application, the false alarm rate can be reduced to less than 5%. If only five sensors are used without spatial interpolation to construct a continuous field, the obstacle avoidance system can only make a binary decision of "turn left" or "turn right" based on the distance in five discrete directions. When positioned between two detection directions, it is prone to misjudgment. Simulation tests show that its obstacle avoidance success rate for slender obstacles (such as cables and railings) with a width of less than 20 centimeters is less than 70%. However, after adopting the virtual radar matrix interpolation of this application, the system can identify the continuous outline of the obstacle, and the obstacle avoidance success rate for the same type of obstacle is increased to more than 95%. While ensuring low cost (using only five general ultrasonic sensors), it also solves the contradiction between "spatial continuous coverage" and "reliable signal acquisition" of obstacles in inland waterway scenarios, and achieves significant technical progress.

[0039] S04: Perform dynamic threat assessment, extract the obstacle center distance, relative approach rate, and azimuth deviation relative to the ship's heading from the obstacle detection field distribution map, and obtain a quantitative comprehensive threat assessment coefficient value through nonlinear weighted formula.

[0040] It is understandable that the calculation process of the comprehensive threat assessment coefficient includes: extracting the reciprocal of the obstacle distance as the first weighting factor; extracting the relative approach speed as the second weighting factor; extracting the obstacle azimuth deviation as the third weighting factor; and performing a nonlinear summation operation on the first, second, and third weighting factors; wherein, the comprehensive threat assessment coefficient is positively correlated with the degree of collision risk.

[0041] Specifically, the reciprocal of the obstacle distance is extracted as the first weighting factor, reflecting the objective law that "the closer the distance, the exponentially increasing the danger." The relative approach speed is extracted as the second weighting factor, endowing the system with dynamic predictive capabilities, enabling it to react in advance to moving obstacles. The azimuth deviation is extracted as the third weighting factor, allowing the system to distinguish the priority of threats from the front and sides. Through the nonlinear weighted synthesis of these three factors, this application establishes a quantitative collision probability model, upgrading obstacle avoidance decision-making from simple threshold judgment to risk assessment with dynamic predictive attributes. The weighting coefficients K1, K2, and K3 in the nonlinear weighting formula are not arbitrarily set, but derived based on statistical analysis of historical inland waterway vessel collision accident data and simulation results of numerous obstacle avoidance scenarios. Specifically, through analysis of hundreds of small inland waterway vessel collision cases, it was found that the collision risk has the most significant relationship with the reciprocal of the distance; the risk increases sharply when the distance is less than twice the vessel length (approximately 2.4 meters). Therefore, K1 is set to 100, so that within a 4-meter range, the contribution of the distance term to the threat coefficient ranges from 0 to 25. The weight K2 for the relative speed factor is set to 10, based on calculations of the ship's braking distance. At typical speeds (1 meter per second), the risk increment from a relative approach speed of 0.1 meters per second is equivalent to the risk increment from a 1-meter reduction in distance. The weight K3 for the azimuth deviation is set to 0.5, determined through simulation experiments. This ensures that the threat posed by an obstacle directly ahead (zero-degree deviation) is approximately 45 threat coefficient units higher than that of an obstacle at a 90-degree lateral angle. This difference is sufficient to drive the system to prioritize threats directly ahead while not completely ignoring close-range risks from the sides. Experimental data shows that using this set of weights, in simulated narrow channel encounter scenarios, the system's collision avoidance success rate is increased by 22% compared to the equal-weight scheme, while the false obstacle avoidance rate is reduced by 15%.

[0042] In this embodiment, a lateral distance balance control process is also included, which includes: extracting the ranging deviation of ultrasonic sensor nodes at positive 90-degree lateral positions and negative 90-degree lateral positions in real time; calculating the difference in distance to obstacles on both sides of the hull in response to the unmanned vessel being in berthing mode or narrow waterway passage mode; and inputting the difference into a lateral drift compensation algorithm to generate a lateral dynamic compensation command for adjusting the berthing attitude.

[0043] Specifically, for special scenarios common in inland waterway navigation, such as berthing at docks or passing through narrow sluice gates and bridge openings, this application also includes a lateral distance balance control process. When the unmanned vessel enters berthing mode or narrow waterway passage mode through upper computer commands or environmental recognition (such as the distance between the two sides being less than two meters and no threat ahead), this process is activated. The system extracts the distance values ​​of ultrasonic sensor nodes at positive 90-degree lateral position and negative 90-degree lateral position in real time, which are recorded as port side distance and starboard side distance, respectively, in meters. The difference in distances to obstacles on both sides of the hull is calculated, i.e., port side distance minus starboard side distance. This difference is input into the lateral drift compensation algorithm, which is actually an independent proportional controller: lateral power compensation command = lateral compensation proportional coefficient * distance difference, where the lateral compensation proportional coefficient can be set to 10. If the difference is positive (port distance greater than starboard distance, i.e., the hull is veering to starboard), a compensation command is generated to slightly accelerate the port thruster or slightly decelerate the starboard thruster, producing a small lateral moment to the left. This causes the hull to adjust slightly to the left, thereby reducing the distance difference between the two sides and bringing the hull closer to the center of the channel or parallel to the dock. This process, through active adjustment of lateral power, significantly improves the lateral distance maintenance accuracy of the unmanned surface vessel when berthing and navigating restricted channels.

[0044] In this embodiment of the application, a dynamic clutter filtering process may also be included, which includes: extracting time-domain features from a two-dimensional heat map; identifying and filtering transient noise with a duration less than a preset time threshold in the ranging data based on the frequency characteristics of the inland river wave motion; and using a median filtering algorithm to process high-frequency artifacts in the detection field distribution map to eliminate false triggering signals caused by water splashing.

[0045] In the process of constructing the virtual radar matrix and threat assessment, in order to improve the reliability of the system in the real water environment, a dynamic clutter filtering process is also included. Due to wind and waves and the navigation of ships, the inland water surface will generate periodic waves, and water splash will also generate instantaneous ultrasonic echoes. These non-obstacle echoes will act as noise interference to the virtual radar matrix. The dynamic clutter filtering process specifically includes: the system extracts time-domain features from a continuous two-dimensional heatmap, that is, it tracks the change sequence of distance values ​​of each grid cell within 10 consecutive operating cycles (i.e., 1 second). Based on the frequency characteristics of inland river wave motion (usually below 2 Hz), the system identifies and filters out transient noise points with a duration less than a preset time threshold (e.g., 200 milliseconds). These short-term changes are usually caused by splashing water. In addition, before filling the original ranging signal into the virtual radar matrix, the system first processes the original distance values ​​of each sensor channel for 5 consecutive cycles using a median filtering algorithm. That is, the median value after sorting these 5 values ​​by size is taken as the effective output of that cycle. This median filtering algorithm can effectively eliminate high-frequency artifacts and outliers caused by electrical interference or random reflections. By combining time-domain feature analysis and median filtering, transient noise caused by splashing water and clutter is effectively filtered out, significantly improving the system's false alarm resistance in harsh water environments.

[0046] In this embodiment of the application, to ensure the stability of ranging accuracy under different climatic conditions, an environmental reference compensation process is also included. The environmental reference compensation process includes: collecting real-time atmospheric temperature data and air humidity data; calculating the real-time sound speed value under the current atmospheric environment based on the sound speed propagation correction formula; and using the real-time sound speed value to perform distance compensation correction on the original ranging signal to eliminate the spatial positioning deviation of the virtual radar matrix caused by changes in ambient temperature and humidity.

[0047] Understandably, the shipboard main control unit is connected to digital temperature and humidity sensors to collect real-time atmospheric temperature and humidity data. The temperature measurement range is -40°C to 80°C, with an accuracy of ±0.5°C; the humidity measurement range is 0 to 100% relative humidity, with an accuracy of ±2%. The speed of sound in air is affected by temperature and humidity, and the correction formula is: Real-time speed of sound (meters per second) = 331.3 + 0.606 * Temperature (°C) + 0.0124 * Relative humidity (percentage). At the beginning of each acquisition cycle, the system reads the data from the temperature and humidity sensors, substitutes it into the above formula to calculate the real-time speed of sound under the current environment. When calculating the original ranging signal, the fixed speed of sound of 340 meters per second is no longer used; instead, this real-time speed of sound value is used for distance calculation, eliminating the drift of the sound speed reference caused by seasonal changes or weather variations, and ensuring the absolute accuracy of the virtual radar matrix in different environments.

[0048] S05: Implement obstacle avoidance execution control, responding to the comprehensive threat assessment coefficient value to trigger obstacle avoidance action, and using a cascaded proportional-integral-derivative controller to convert the comprehensive threat assessment coefficient value into the corresponding track deviation correction amount and propulsion power adjustment amount. The propulsion power adjustment amount includes the forward speed adjustment command for the unmanned vessel propulsion and the reverse braking command in emergency obstacle avoidance conditions.

[0049] Understandably, the implementation process of a cascaded proportional-integral-derivative controller includes: establishing a closed-loop control structure with the heading angle deviation as the outer loop and the thruster speed as the inner loop; dynamically adjusting the proportional coefficient, integral coefficient, and derivative coefficient of the controller according to the magnitude of the comprehensive threat assessment coefficient; and prioritizing the correction of the unmanned vessel's heading by adjusting the rudder angle while meeting the preset safe distance conditions.

[0050] Specifically, the implementation process of the cascaded proportional-integral-derivative (PID) controller includes: establishing a closed-loop control structure with the heading angle deviation as the outer loop and the thruster speed as the inner loop. The input to the outer loop is the deviation between the desired heading angle and the actual heading angle (measured by an electronic compass). However, the desired heading angle is not given by the global path point, but is dynamically generated by the comprehensive threat assessment coefficient value. Specifically, the system sets a threat coefficient threshold, for example, 50. When the comprehensive threat assessment coefficient value is lower than this threshold, the current environment is considered safe, the obstacle avoidance controller does not intervene, and the unmanned surface vessel travels according to the preset heading. When the comprehensive threat assessment coefficient value exceeds this threshold, obstacle avoidance action is triggered. At this time, the system first generates a temporary desired heading angle offset based on the azimuth deviation of the obstacle center. For example, if the obstacle is 30 degrees to the left front, a desired heading angle command of turning 30 degrees to the right is generated. This heading angle deviation command serves as the input to the outer loop PID controller. The proportional, integral, and derivative coefficients of the cascaded proportional-integral-derivative (PID) controller can be dynamically adjusted based on the magnitude of the comprehensive threat assessment coefficient. The specific strategy was determined through extensive pool experiments and simulation tests to balance response speed and stability under different threat levels. The specific parameter mapping relationship for dynamic adjustment is as follows: When the comprehensive threat assessment coefficient is between 50 and 75, it is defined as a low threat level. At this time, the proportional coefficient KP of the outer-loop PID controller is set to 1.5, the integral coefficient KI to 0.05, and the derivative coefficient KD to 0.2. When the threat coefficient is between 75 and 125, it is defined as a medium threat level. The proportional coefficient KP is adjusted to 0.5, the integral coefficient KI to 0.02, and the derivative coefficient KD to 0.5 to enhance the system's rapid response capability and suppress integral saturation. When the threat coefficient is greater than 125, it is defined as a high threat level. The proportional coefficient KP is further increased to 4, the integral coefficient KI is decreased to 0.01 (or set to zero), and the derivative coefficient KD is increased to 1 to obtain the fastest obstacle avoidance action while avoiding overshoot. Furthermore, the functional relationship of the parameters continuously changing with the threat coefficient can also be described as follows: KP = 1.0 + 0.02 * (threat coefficient value - 50), taking the value of 1 when the threat coefficient value is less than 50. This linear relationship makes the control strength increase smoothly with the threat; KI = 0.1 / (1 + 0.01 * (threat coefficient value - 50)), showing an inverse relationship, weakening the integral action at high threats to prevent cumulative errors from causing control divergence; KD = 0.1 + 0.005 * (threat coefficient value - 50), linearly increasing the derivative action to improve system damping. The above parameter adjustment strategy ensures that the controller can adapt to obstacle avoidance scenarios with different levels of urgency.The outer loop controller outputs the desired rudder angle command, which is sent to the servo actuator. Simultaneously, the inner loop controller begins operation, with its input being the deviation between the desired propeller speed and the actual speed. During obstacle avoidance, propulsion power adjustment includes two types: one is a forward speed adjustment command for the unmanned vessel's propellers, such as using differential control (decelerating on the port side and accelerating on the starboard side to achieve a right turn) to assist steering during turns; the other is in emergency obstacle avoidance situations, when the threat coefficient is extremely high (e.g., greater than 150) and the obstacle is very close, the system will issue a reverse braking command, which controls the two propeller motors to reverse briefly to provide emergency braking force and quickly reduce the vessel's speed.

[0051] In this embodiment of the application, a heading smoothing process is also included, specifically including: configuring a low-pass filter at the output of the cascaded proportional-integral-derivative controller; setting the dead zone range for heading correction, and keeping the current thruster state unchanged when the track deviation correction amount is less than the preset minimum dead zone value; and eliminating the heading oscillation phenomenon generated during continuous obstacle avoidance by limiting the maximum slope of the rudder angle change rate.

[0052] Specifically, to avoid severe oscillations or serpentine trajectories caused by sudden changes in obstacle avoidance commands, the method in this application also includes a course smoothing process. This process specifically includes configuring a first-order low-pass filter at the output of the cascaded proportional-integral-derivative controller, with a cutoff frequency set to 5 Hz, to filter out high-frequency jitter components in the control commands. The system sets a dead zone range for course correction, with a dead zone threshold set to ±0.5 degrees. This means that when the absolute value of the calculated track deviation correction (i.e., the desired change in heading angle) is less than 0.5 degrees, the controller will ignore this small deviation and maintain the current rudder angle and propeller state unchanged. This avoids meaningless frequent fine-tuning of the control system when there is no threat or the threat is minor, thus improving navigation stability. In addition, the system limits the maximum slope of the rudder angle change rate through software, that is, it stipulates that the change of the rudder angle command within every 0.1-second time interval shall not exceed 5 degrees. This limitation forces the rotation speed of the rudder to be relatively slow. Even if the upper controller outputs a large step command, the actual rudder angle change is a smooth slope process, which effectively eliminates the heading oscillation phenomenon caused by rapid alternation of commands during continuous obstacle avoidance, and ensures the navigation stability of the unmanned vessel under continuous obstacle avoidance conditions.

[0053] This application also discloses an inland waterway unmanned vessel collision avoidance perception and obstacle avoidance control system, including: a memory, a processor, and a computer program stored in the memory and capable of running on the processor. When the processor executes the computer program, it implements the above-described inland waterway unmanned vessel navigation collision avoidance method based on multi-point ultrasonic ranging.

[0054] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional modules is used as an example. In practical applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above. The specific working process of the system, device, and unit described above can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0055] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of modules or units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces, or indirect coupling or communication connection between apparatuses or units, and may be electrical, mechanical, or other forms.

[0056] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0057] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0058] If the integrated unit is implemented as 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 this application, in essence, 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. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) or processor to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory, random access memory, magnetic disks, or optical disks.

[0059] The above are all preferred embodiments of this application and are not intended to limit the scope of protection of this application. Any feature disclosed in this specification (including the abstract and drawings) may be replaced by other equivalent or similar features unless specifically stated otherwise. That is, unless specifically stated otherwise, each feature is only one example of a series of equivalent or similar features.

Claims

1. A collision avoidance method for unmanned surface vessels (USVs) in inland waterways based on multi-point ultrasonic ranging, characterized in that, include: An ultrasonic sensor array is configured, comprising five ultrasonic sensor nodes. The five ultrasonic sensor nodes are respectively installed in a fan-shaped topology at the zero-degree central axis position, at a positive 45-degree oblique position and a negative 45-degree oblique position symmetrically distributed with respect to the central axis, and at a positive 90-degree lateral position and a negative 90-degree lateral position symmetrically distributed with respect to the central axis. The original ranging signals of each ultrasonic sensor node are collected cyclically based on the time-division multiplexing mechanism; A virtual radar matrix is ​​constructed, and the discrete original ranging signals are normalized and mapped and projected into a two-dimensional polar coordinate space model. The virtual ranging values ​​of each detection gap point are calculated by a spatial interpolation algorithm to generate a continuous obstacle detection field distribution map. Perform dynamic threat assessment, extract the obstacle center distance, relative approach rate and azimuth deviation relative to the ship's bow direction from the obstacle detection field distribution map, and obtain the quantified comprehensive threat assessment coefficient value through nonlinear weighted formula; Obstacle avoidance execution control is implemented, and obstacle avoidance actions are triggered in response to the comprehensive threat assessment coefficient value. The comprehensive threat assessment coefficient value is converted into the corresponding track deviation correction amount and propulsion power adjustment amount using a cascaded proportional-integral-derivative controller. The propulsion power adjustment amount includes a forward speed adjustment command for the unmanned vessel's propulsion and a reverse braking command in emergency obstacle avoidance conditions.

2. The collision avoidance method for inland waterway unmanned vessels based on multi-point ultrasonic ranging according to claim 1, characterized in that, The process of constructing the virtual radar matrix includes: A spatial grid model in polar coordinates is established, wherein the angle step size of the spatial grid model is configured to be 1 to 5 degrees; The original ranging signal is marked on the corresponding zero, ±45, and ±90 degree radial axes; The depth values ​​of empty grid cells between adjacent radial axes are calculated using a bilinear interpolation algorithm; A continuous ranging envelope covering a 180-degree area in front of the unmanned vessel is generated, forming a two-dimensional heat map reflecting the spatial morphology of obstacles.

3. The collision avoidance method for inland waterway unmanned vessels based on multi-point ultrasonic ranging according to claim 2, characterized in that, The method further includes a dynamic clutter filtering process, which includes: Temporal features are extracted from the two-dimensional heatmap; Based on the frequency characteristics of inland river wave motion, transient noise with a duration less than a preset time threshold is identified and filtered out in ranging data. The high-frequency artifacts in the detection field distribution map are processed using a median filtering algorithm to eliminate false triggering signals caused by water splashing.

4. The collision avoidance method for inland waterway unmanned vessels based on multi-point ultrasonic ranging according to claim 1, characterized in that, The process of cyclic acquisition based on the time-division multiplexing mechanism includes: Configure a clock synchronization controller to divide the five ultrasonic sensor nodes into non-interfering transmission groups; The ultrasonic pulse trigger command is sent cyclically in the order of zero-degree center position, ±45-degree position, and ±90-degree position; A preset time delay interval is set between the trigger commands of adjacent groups, wherein the preset time delay interval is configured to be greater than twice the round-trip propagation time of the ultrasonic wave within a preset maximum range.

5. The collision avoidance method for unmanned inland waterway vessels based on multi-point ultrasonic ranging according to claim 1, characterized in that, The calculation process for the comprehensive threat assessment coefficient value includes: Extract the reciprocal of the obstacle distance as the first weighting factor; The relative approximation velocity is extracted as the second weighting factor; The obstacle azimuth deviation is extracted as the third weighting factor; Perform a nonlinear summation operation on the first weighting factor, the second weighting factor, and the third weighting factor; The comprehensive threat assessment coefficient is positively correlated with the degree of collision risk.

6. The collision avoidance method for inland waterway unmanned vessels based on multi-point ultrasonic ranging according to claim 5, characterized in that, It also includes a lateral distance balance control process, specifically including: The ranging deviation of the ultrasonic sensor node at the positive 90-degree lateral position and the negative 90-degree lateral position is extracted in real time. In response to the unmanned vessel being in berthing mode or narrow waterway passage mode, the difference in distance to obstacles on both sides of the hull is calculated; The difference is input into the lateral drift compensation algorithm to generate lateral dynamic compensation commands for adjusting the berthing attitude.

7. The collision avoidance method for unmanned inland waterway vessels based on multi-point ultrasonic ranging according to claim 1, characterized in that, The implementation process of the cascaded proportional-integral-derivative controller includes: Establish a closed-loop control structure with heading angle deviation as the outer loop and thruster speed as the inner loop; Based on the magnitude of the comprehensive threat assessment coefficient, the proportional coefficient, integral coefficient, and derivative coefficient of the controller are dynamically adjusted. Under the condition of meeting the preset safe distance, the unmanned vessel's course is corrected by adjusting the rudder angle first.

8. The collision avoidance method for unmanned inland waterway vessels based on multi-point ultrasonic ranging according to claim 7, characterized in that, It also includes a course smoothing process, specifically including: A low-pass filter is configured at the output of the cascaded proportional-integral-derivative controller. Set the dead zone range for heading correction. When the track deviation correction amount is less than the preset minimum dead zone value, keep the current thruster state unchanged. By limiting the maximum slope of the rudder angle change rate, the heading oscillation phenomenon generated during continuous obstacle avoidance is eliminated.

9. The collision avoidance method for unmanned inland waterway vessels based on multi-point ultrasonic ranging according to claim 1, characterized in that, It also includes the environmental baseline compensation process, specifically including: Real-time acquisition of atmospheric temperature and humidity data; Calculate the real-time sound speed value under the current atmospheric environment based on the sound speed propagation correction formula; The original ranging signal is corrected by distance compensation using the real-time sound speed value to eliminate spatial positioning deviation of the virtual radar matrix caused by changes in ambient temperature and humidity.

10. A collision avoidance system for unmanned inland waterway vessels based on multi-point ultrasonic ranging, characterized in that, include: Memory; processor; as well as A computer program stored in the memory and executable on the processor, wherein the processor, when executing the computer program, implements the collision avoidance method for inland unmanned vessels based on multi-point ultrasonic ranging as described in any one of claims 1 to 9.

Images