Quantitative method of ship navigation risk under three-dimensional time-delay speed obstacle of polar sea ice

Abstract

The application provides a ship navigation risk quantification method under a polar sea ice three-dimensional-time delay speed obstacle, and the method comprises the following steps: collecting local sensor data of a ship to form a ship motion state and an ice block motion set; constructing a three-dimensional-time delay speed obstacle model TDIVO³D to process the ship motion state and the ice block motion set, and generating two three-dimensional binary mask graphs, including a speed obstacle domain and a feasible speed domain; mapping the speed obstacle domain and the feasible speed domain into a continuous quantification index sensitive to polar navigation risk; designing a target function for risk avoidance or minimum angle change according to the speed obstacle domain, the feasible speed domain and the continuous quantification index, calculating the target function, selecting a speed vector corresponding to a minimum value as an optimal avoidance speed vector, and performing navigation risk avoidance on the ship based on the optimal avoidance speed vector. The application has outstanding practical value and technical frontiers in the polar environment with much ice, many ships and high uncertainty.

Description

TECHNICAL FIELD

[0001] The application belongs to the field of polar ship navigation risk quantification, and particularly relates to a ship navigation risk quantification method under a three-dimensional-time delay velocity obstacle of polar sea ice. BACKGROUND

[0002] With the continuous extension of the Arctic navigation season, more and more ocean-going ships attempt to cross the polar sea area with dense floating ice and bad weather. However, the polar environment has high dynamic and asymmetric characteristics: ice blocks can drift at a non-uniform speed under the joint driving of wind and current, and the ice keel under the ice blocks forms a typical "unobservable dangerous source" in the visual blind area, which can still cause structural damage under the shallow-draft ship. In addition, the low temperature in the polar region can cause a response delay of the propulsion system of up to 5-10 seconds, causing an unstable control phenomenon of "maneuvering and speed response decoupling". At the same time, the ionosphere is severely disturbed under the polar night conditions, which aggravates the GNSS position drift, and the positioning accuracy relied on by the conventional trajectory planning method is invalid, further causing the risk of misjudgment and delayed judgment. Under this background, the traditional two-dimensional collision avoidance index based on DCPA / TCPA is insufficient in stability due to its inability to cover the vertical structural risk and time response delay in the scene of complex ice-ship interaction and severe navigation accuracy fluctuation. At present, the improved velocity obstacle method also stays at the level of two-dimensional velocity field construction, lacking the system modeling of propulsion power delay, ice body rotation influence and multi-time and space factor interaction. In actual operation, in order to avoid collision, the ship is often forced to slow down or detour, thereby losing a lot of navigation efficiency.

[0003] Therefore, there is an urgent need for a complete method system that can integrate three-dimensional perception, multi-scale motion prediction, time delay control and risk executable quantification, and can adapt to the high uncertainty in the polar environment to support automatic decision and closed-loop control of the collision avoidance system. SUMMARY

[0004] The purpose of the present application is to propose a ship navigation risk quantification method under a three-dimensional-time delay velocity obstacle of polar sea ice, which solves the deficiencies in the existing navigation risk quantification from sensor information integration, dynamic modeling, risk quantification, velocity search to instruction execution.

[0005] In order to achieve the above purpose, the present application provides a ship navigation risk quantification method under a three-dimensional-time delay velocity obstacle of polar sea ice, which comprises:

[0006] S100, collecting ship local sensor data to form a ship motion state and an ice block motion set;

[0007] S200, constructing a three-dimensional-time delay velocity obstacle model TDIVO³D processing the ship motion state and the ice block motion set to generate two three-dimensional binary mask maps, including a velocity obstacle domain and a feasible velocity domain;

[0008] S300, mapping the speed obstacle domain and the feasible speed domain as a continuous quantified index sensitive to polar navigation risk;

[0009] S400, designing a target function with minimum collision avoidance risk or angle change based on the speed obstacle domain and the feasible speed domain and the continuous quantified index, calculating the target function, and selecting a speed vector corresponding to a minimum value as an optimal avoidance speed vector;

[0010] S500, performing navigation risk avoidance for the ship based on the optimal avoidance speed vector.

[0011] Further, the ship local sensor data includes a sequence of space-borne synthetic aperture radar images, a rough ship attitude provided by a GNSS / IMU combined navigation, a sailing speed and a heading information output by a ship-borne AIS system, and a green laser compound point cloud of a multi-beam sonar of a ship bow forward view, and the ship local sensor data is uniformly mapped to a local coordinate system with a ship body center as an origin.

[0012] Further, the ship local sensor data is collected to form a ship motion state and an ice block motion set, specifically:

[0013] According to a preset ship local data fusion module, ship local sensor data is received;

[0014] The ship local sensor data is preprocessed to form a ship motion state and an ice block motion set; wherein the preprocessing includes coherent difference processing, iterative nearest point rigid body matching, and extended Kalman filter processing;

[0015] The ship motion state includes a fusion state vector of a current frame of the ship and a ship center path; and the ice block motion set includes a fusion state vector of a current frame of the ice block and an ice block center path.

[0016] Further, the three-dimensional-time-delay velocity obstacle model TDIVO³D processes the ship motion state and the ice block motion set to generate two three-dimensional binary mask maps, including a speed obstacle domain and a feasible speed domain, specifically:

[0017] The maximum propulsion rate and the maximum turning angle rate of the ship are obtained, and a discretized two-dimensional speed-heading grid is constructed;

[0018] An Euler forward integration is performed on each candidate speed vector in the two-dimensional speed-heading grid to obtain a predicted spatial path of the ship within 60 seconds in the future, and a time delay correction amount is introduced to delay the entire path by the time delay correction amount, to obtain a feasible speed domain after delay compensation;

[0019] According to the hull itself safety domain in the horizontal direction and the ice block circumscribed radius, a comprehensive safety radius of each ice block is constructed;

[0020] A parallel Minkowski sum calculation is performed on the GPU for each candidate velocity vector to obtain a velocity obstacle.

[0021] Further, the parallel Minkowski sum calculation performed on the GPU for each candidate velocity vector to obtain a velocity obstacle is specifically:

[0022] For each ice block, it is judged whether the predicted path of the ship will enter the comprehensive safety radius corresponding to the ice block within a sliding time window If the ship advances at a speed , the predicted path will enter the comprehensive safety radius; if the predicted path is less than the comprehensive safety radius, the speed vector is marked as a dangerous point and is included in the velocity obstacle.

[0023] Further, the mapping of the velocity obstacle and the feasible speed domain into a continuous quantitative index sensitive to polar navigation risk is specifically:

[0024] The sum operation of the two mask voxel sets of the velocity obstacle and the feasible speed domain is completed in parallel on the GPU to obtain a volume ratio; in an ideal navigation state, the volume ratio is less than about 0.1; once the volume ratio exceeds 0.3, it indicates that there is no safe and feasible solution for the ship in the feasible speed domain;

[0025] When in the navigation state with a volume ratio less than about 0.1, the positioning covariance of the velocity obstacle is converted into a scalar , that is, the trace of the positioning covariance is taken as the overall error index, and a continuous quantitative index is calculated using the following exponential function :

[0026] ;

[0027] Wherein, the steepness of the control risk growth, is the decision risk benchmark point, that is multiplied by the uncertainty and exceeds , the risk begins to rise significantly; is a mapping scale, which is used to normalize the covariance dimension to a dimensionless volume ratio matching interval;

[0028] Further, the exponential function is used to ensure that even if there is a certain degree of covariance, the risk will not be amplified to an unusable state when the mapping scale is small; but when the mapping scale approaches And the covariance increases, the continuous quantization index Will quickly approach to 1, prompting the system or the driver that the current speed domain is almost unable to meet the safety traffic requirements.

[0029] Further, the target function designed to minimize the collision risk or the angle change according to the speed obstacle domain and the feasible speed domain and the continuous quantization index, the target function is calculated, and the speed vector corresponding to the minimum value is selected as the optimal avoidance speed vector, specifically:

[0030] The speed obstacle domain is removed from the feasible speed domain to obtain a speed safety sub-domain, wherein the speed safety sub-domain includes a plurality of candidate speed vectors; the candidate speed vector includes speed and angle;

[0031] According to the candidate speed vector and the continuous quantization index The target function of minimizing the collision risk or the angle change is designed, and a response hysteresis cost function is introduced into the target function, indicating the required control response time of the current speed vector, and the optimal avoidance speed vector and the corresponding minimum generation value are calculated through the target function;

[0032] Wherein, the response hysteresis cost function is a dynamic function, in the polar region low temperature environment, according to the low temperature propulsion response time, rudder angle response time, current maximum acceleration and maximum turning ability of the ship and the candidate speed vector.

[0033] Further, the ship is navigated based on the optimal avoidance speed vector, specifically:

[0034] The optimal avoidance speed vector is converted into a standard control instruction format;

[0035] According to the standard control instruction format combined with the preset calibration lookup table data, the corresponding propulsion power setting value and rudder angle target are obtained by interpolation solution;

[0036] The propulsion power setting value and rudder angle target are executed, and the ship state after actual execution is read at

[0037] Based on the ship state combined with the optimal avoidance speed vector, the execution error is calculated;

[0038] Based on the execution error, the resolution of the feasible speed search is adjusted.

[0039] Further, the standard control instruction format is packaged in NMEA or IEC61162-450 encoding format, and is sent to the ship electric control navigation control module or intelligent course keeping system through CAN bus.

[0040] ​The beneficial technical effects of the present application are at least the following:

[0041] The present application proposes a closed-loop risk quantification and avoidance instruction generation method for polar navigation environment, constructs the dynamic state of the ship and sea ice in the unified coordinate system based on the multi-sensor asynchronous sensing mechanism, and fuses the propulsion delay, the spin ice body expansion domain and the vertical buffer of the ice under the keel in the three-dimensional velocity space, and first constructs a three-dimensional-time delay velocity obstacle model containing time delay and vertical dimension. The velocity obstacle volume and the feasible velocity volume output by the model are mapped into a risk index through the normalized volume ratio, and a multi-objective optimization function containing a delay response penalty term is further introduced in the risk field, so that the risk assessment and optimal avoidance decision are integrated. On this basis, the system constructs a standardized closed-loop interface from the velocity instruction to the ship control execution unit, all optimal velocity vectors can be converted into actual propulsion power and rudder angle instructions, and the sampling accuracy and prediction parameters are automatically adjusted through state feedback error, forming a 10Hz real-time cycle risk-instruction closed loop. This scheme is fully connected from sensor information integration, dynamic modeling, risk quantification, velocity search to instruction execution, has high engineering reproducibility, can be adapted to different ship types and control interfaces, and has outstanding practical value and technical frontier in the polar multi-ice, multi-ship and high uncertainty environment. BRIEF DESCRIPTION OF DRAWINGS

[0042] The present application is further illustrated by the accompanying drawings, but the embodiments in the drawings do not constitute any limitation on the present application. For ordinary skilled in the art, other drawings can be obtained without creative labor on the basis of the following drawings.

[0043] Figure 1 The present application is a polar sea ice three-dimensional-time delay velocity obstacle ship navigation risk quantification method flow chart.

[0044] Figure 2 The present application is a polar two-ship encounter position diagram.

[0045] Figure 3 The present application is a DIVO model diagram.

[0046] Figure 4 The present application is a polar two-ship encounter CRI, velocity, relative distance, DCPA, TCPA diagram. DETAILED DESCRIPTION

[0047] The embodiments of the present application are described in detail below, and examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference signs represent the same or similar elements or elements having the same or similar functions throughout. The embodiments described below by referring to the drawings are exemplary and are only used to explain the present application, and cannot be understood as a limitation on the present application.

[0048] like Figure 1 As shown in the embodiment of the present invention, the method for quantifying ship navigation risks under three-dimensional-time-delay velocity obstacles in polar sea ice includes the following steps S100-S500:

[0049] S100: Collects local sensor data from the ship to form a set of ship motion status and ice block motion data.

[0050] Specifically, in this step, the system synchronously receives four types of key sensor information through the ship's local data fusion module, namely: spaceborne synthetic aperture radar image sequences. (Sampling frequency of 2Hz), the coarse ship attitude provided by GNSS / IMU integrated navigation. (10Hz) Shipborne AIS system outputs speed and heading information (1Hz), and the forward-looking multibeam sonar-green laser composite point cloud at the bow. (5Hz). All data is timestamped via PTP (PrecisionTimeProtocol) after access and uniformly mapped to the local coordinate system with the ship's center of mass as the origin. This is to ensure spatial consistency and temporal alignment of data from different sensors.

[0051] In polar navigation environments, severe ionospheric disturbances at high latitudes can cause GNSS signals to experience instantaneous jumps or shifts, directly affecting the reliability of ship positioning. To improve pose accuracy, the system constructs a sparse graph optimization model every 5 seconds within a sliding window, using SAR image sequences... The texture feature points in the data are used as constraints, and they are compared with... Fusion calculation to correct pose And simultaneously output the residual covariance matrix. This covariance will then be used to estimate the dynamic expansion of the ice body profile, thereby enhancing the robustness of sea ice detection.

[0052] In terms of obstacle modeling, the system first performs... Coherent differential processing is performed on two adjacent frames to highlight strong scattering objects whose positions change within the time interval, i.e., moving obstacles such as drifting ice blocks / ships. Then, the sonar-laser point cloud is... Projecting the image onto the radar image plane improves the geometric consistency between the outlines in the image and the actual ice blocks, especially for morphological completion of sub-ice keels that are partially obscured by the sea surface but have significant vertical volume. After connected component analysis, the system assigns a unique number to each closed ice block region. .

[0053] like Figure 2As shown, this represents the initial state in a typical two-ship encounter scenario. The blue and orange arrows represent the initial positions and headings of the two targets (S1 and S2), respectively, indicating that they form a head-on encounter relationship.

[0054] Furthermore, this applies to moving obstacles such as ice blocks / ships with the same identification number. The system performs GPU-based Iterative Closest Point (ICP) rigid body matching in two consecutive frame point clouds to extract the ice block's position. Translational velocity in the plane and rotational angular velocity This velocity information will be used to calculate the ice block's center of mass at any given moment. Path:

[0055] ;

[0056] In this formula, This indicates the initial centroid position of a moving obstacle such as an ice block or ship when it is first detected. These represent the lateral and longitudinal translational velocities of moving obstacles such as ice blocks / ships. This represents the rotational angular velocity; all values ​​are output by the GPU-ICP module, ensuring high real-time performance and stability. The integral form allows for the subsequent reproduction of the ice block's trajectory within any prediction time window, forming the basis for building dynamic velocity obstacle models.

[0057] For modeling the ship's own condition, it will start from Extracted speed and heading, and attitude-corrected velocity components The fused state vector of the current frame is obtained by inputting both into the extended Kalman filter. ,in Indicates the ship's speed. Indicates the current heading. Combined with the heading unit vector. The path of the ship's center of mass can be obtained by integrating the following formula:

[0058] ;

[0059] in This represents the reference centroid of the ship at the moment of system startup or the initial stage of the voyage. This integral is used to continuously generate the ship's motion path and, in conjunction with the ice block trajectory, participates in the construction of the velocity obstacle volume for the next stage.

[0060] In summary, this step outputs two sets of key data:

[0061] Ice Cube Collection ,in This describes both translation and rotation.

[0062] Ship status The current sailing speed state and the trajectory estimation result.

[0063] Both groups of data are located in the local coordinate system and have been normalized, which can be directly used as input for constructing the speed obstacle model, without the need for space-time conversion or unit normalization, effectively improving the subsequent modeling efficiency and overall system response rate.

[0064] S200, construct a three-dimensional-time delay speed obstacle model TDIVO³D process the ship motion state and ice movement set, generate two three-dimensional binary mask images, including the speed obstacle domain and the feasible speed domain.

[0065] Specifically, as shown in Figure 2 , Figure 3 In this step, the system constructs a three-dimensional-time delay speed obstacle model TDIVO³D based on the ship motion state and the ice movement set output by S100 for dynamic collision avoidance analysis. All variables have been mapped to the local ship coordinate system and normalized to ensure that the speed space and obstacle space are comparable in numerical scale. The key to this step is to improve the traditional DIVO model in the light of the particularity of the polar environment, including the response lag of the propulsion system, the vertical collision risk caused by the ice under the keel, and the asymmetric influence of ice rotation. The feasible speed domain and the speed obstacle domain are output to meet the needs of polar navigation, providing accurate, real-time, and differentiable input for the next step of risk volume ratio calculation.

[0066] Further, first, according to the limit capacity of the ship power system (including the maximum propulsion speed and the maximum turning angle rate), a discretized two-dimensional speed- heading grid is constructed, with a resolution of kn× For each candidate speed vector in the grid, Euler forward integration is used to perform state deduction for a 60-second time window, i.e., to predict which spatial paths the ship will pass through in the future s if it travels at that speed. To consider the response lag of the propulsion system in the polar low-temperature environment, a lag correction amount is introduced (obtained from the environmental temperature-rotation speed response characteristic curve lookup table provided by the shipyard), and all paths are delayed by time to form a feasible speed domain after lag compensation. This domain represents the set of all possible actual speed combinations within the next 60 seconds under the current ship power conditions.

[0067] Next, to depict the dynamic influence of ice obstacles on the ship's speed domain, each ice Construct its comprehensive safety radius This is to quantify the extent of the ice block's impact on the collision risk to the ship. This radius includes not only the horizontal safety zone of the ship itself. Circumradius of the ice cube It also introduces the following two extensions specific to polar scenarios:

[0068] ;

[0069] in, It is the vertical safety margin mapped from the ship's draft and the maximum depth of the keel under ice as inverted by sonar, used to extend the collision envelope in the vertical direction; Let be the angular velocity of the ice cube's rotation. This is a rotational penalty coefficient related to ice thickness and ice stiffness, used to amplify the collision threat posed by high-speed rotating ice to ships. This model structure, based on the traditional DIVO model, introduces rotational effects and vertical compensation terms for the first time, reflecting the asymmetric and highly coupled dynamic characteristics of navigation risks in polar ice regions.

[0070] Subsequently, based on the above definition, the system performs processing on the GPU for each candidate velocity vector. Perform parallel convolution computation: for each piece of ice Determine within the sliding time window Inside, if the ship is at speed (Considering hysteresis) Moving forward, will its predicted path enter the comprehensive safety envelope corresponding to the ice block? If the following conditions are met, the velocity vector is marked as a danger point and included in the obstacle domain. :

[0071] ;

[0072] In this formula, To take into account the time lag in predicting the future position of ships, For ice cubes exist The centroid position at any given time is provided by the integration path of S100. This determination is spatially represented by Minkowski summation and computationally achieved through voxel mask convolution. With a 256×128 grid configuration, the average computation time is no more than 18ms, meeting the real-time processing requirements of 10Hz.

[0073] After convolution, the system generates two 3D binary mask images: one representing... That is, the maneuverable velocity vector under all constrained conditions; a representation These are all combinations of speeds that would pose a collision risk within the current time period. These two masks will be directly used for volume ratio calculation and risk index generation in the next step, and will be cached in GPU memory during this step to avoid repeated readings and reconstructions, thus improving overall efficiency.

[0074] Output: Feasible velocity domain after propulsion hysteresis correction Speed ​​barrier zone This includes the risk of vertical and rotational expansion.

[0075] S300, Map the speed barrier domain and feasible speed domain to a continuous quantitative index that is sensitive to the risks of polar navigation.

[0076] Specifically, this step is responsible for the speed obstacle zone. With feasible velocity domain This is mapped to a continuous quantitative indicator that is operable, interpretable, and sensitive to the risks of polar navigation. In polar navigation environments, traditional two-dimensional collision risk assessment methods (such as DCPA / TCPA) cannot cover the uncertainties unique to high latitudes, such as multi-directional drift, sea ice rotation, and GNSS signal drift. This step, based on three-dimensional velocity space, constructs a nonlinear risk mapping function by integrating the proportion of velocity obstacles and the covariance of positioning drift through voxel modeling and uncertainty adjustment. This makes the risk representation more consistent with the operational logic of ice-covered areas and the judgment habits of crew members.

[0077] First, at the voxel level, the volume ratio of the obstacle velocity domain to the overall velocity domain is obtained through mask statistics. The summation operation of two mask voxel sets is performed in parallel on the GPU side, denoted as follows: and Calculate the volume ratio :

[0078] ;

[0079] This ratio represents the proportion of feasible speeds "blocked" by current icing conditions and power limitations; a higher ratio indicates a more concentrated collision risk and more restricted speed selection. Under ideal navigation conditions, It is approximately less than 0.1; once it exceeds 0.3, it indicates that there is almost no safe and feasible solution for the ship in the multidimensional velocity domain.

[0080] However, in polar scenarios, the instability of GNSS accuracy can cause a deviation between the current pose estimation and the actual position, resulting in some... The boundary conditions can be misjudged or delayed. Therefore, this invention introduces an uncertainty adjustment coefficient to adjust the positioning covariance. Convert to scalar That is, take the positioning covariance. traces , as the overall error indicator, the normalized risk index is calculated as follows:

[0081] ;

[0082] In the formula, the steepness of the risk growth is controlled, is the decision risk benchmark point (i.e. multiplied by the uncertainty, and when , the risk begins to rise significantly), and is a mapping scale used to normalize the covariance dimension to a dimensionless volume ratio matching interval. This design replaces the "fixed risk threshold" in traditional models, making the risk expression no longer dependent on artificial constants, but able to change in real time with environmental and positioning conditions.

[0083] The exponential function ensures that when is small, even if there is a certain degree of covariance, it will not amplify the risk to an unusable state; but when approaches and the covariance increases, will quickly approach 1, reminding the system or the driver that the current speed domain is almost unable to meet the safety travel requirements. On the bridge display terminal, is mapped in real time to a three-color risk bar (green-yellow-red), which is used in conjunction with the forward-looking AR visualization module for reference by the crew.

[0084] At the same time, to avoid the influence of high-frequency disturbances on risk stability, the invention introduces a sliding window smoothing mechanism for , and the current frame value is taken as the average of the last 3 frames to remove short-period fluctuations. This mechanism ensures that even if the sea ice is short of shock and the speed changes slightly, the risk display remains stable, avoiding misleading the crew.

[0085] The three-dimensional normalized risk index is obtained, which is used by the next decision module; and the original volume ratio can be used for recording and post-processing parameter adjustment, or for providing original comparison for the driving interface.

[0086] S400, according to the degree of obstacle domain and, the feasible speed domain and the continuous quantitative index, design a target function to minimize collision risk or angle change, calculate the target function, and select the speed vector corresponding to the minimum value as the optimal avoidance speed vector.

[0087] Specifically, in this step, the system needs to calculate the three-dimensional risk index output by the previous stage, the current feasible speed domain and the speed obstacle domain , for the current state of the polar ship Select a set of optimal safety speed vectors The vectors not only meet the dynamic collision avoidance requirements, but also minimize the operating cost as much as possible, avoid high-risk routes, and consider the time penalty caused by low-temperature control hysteresis. In polar navigation, the captain pays more attention to the overall energy consumption and response controllability when choosing to avoid action, rather than just whether the avoidance is successful. Therefore, designing a reasonable optimization objective function is the key creative point of this step.

[0088] First, the system eliminates from to obtain the speed safety sub-domain All candidate speed vectors come from this domain. Due to the problems of ice keel, high-density ice, and propulsion response lag in the polar environment, the objective function designed in this step is not only to minimize collision risk or angle change, but also to introduce a combined function that comprehensively considers "risk cost, power cost, and controllability", which is in the form of:

[0089] ;

[0090] Wherein: the first term represents the change in sailing speed, reflecting energy consumption and propulsion response consumption; the second term represents the deviation of the sailing direction, reflecting the controllability of the operation; the third term is the risk index from the previous step, which has time smoothing and reflects the subjective avoidance intention of collision risk;

[0091] The fourth term is the "response hysteresis cost function" first introduced in this step, which has the physical meaning of the response time required to reach the speed vector In the polar low-temperature environment, this time is significantly higher than that in the normal-temperature sea area. The invention defines:

[0092] ;

[0093] Wherein are the low-temperature propulsion response time (such as 5-8 seconds) and the rudder angle response time (such as 3-5 seconds), respectively, which are obtained by power system simulation or static test lookup table; is the current maximum acceleration and maximum turning ability of the ship.

[0094] It can be understood that the innovation point of this design is that is a dynamic function that does not change linearly with the speed vector, but estimates the cost in real time according to the actual response performance of the system in the polar low-temperature environment, which has strong physical significance. It makes the optimal speed search not only selected from "geometric feasibility", but also prioritizes those "fast response, low risk, and stable operation" instructions.

[0095] Among all candidate , the system calculates based on the cost function mentioned above in parallel, and selects the speed vector corresponding to the minimum value as the optimal control variable:

[0096] ;

[0097] This search process is implemented using GPU, and all candidate vectors have been pre-mapped on the structured voxel grid, without the need for dynamic path calculation or boundary collision judgment, ensuring a real-time control refresh rate of 10Hz.

[0098] The optimal avoidance speed vector is obtained: the next step will be issued to the ship navigation control system or the driving assistance module; the minimum cost value : reflects the composite index of operating safety margin and response burden under the current sea ice conditions, which can be used for visualization or subsequent dynamic adjustment of weight parameters.

[0099] This step first introduces the low-temperature propulsion delay cost in an explicit form into the speed optimization function, and combines the continuous risk field from S300, to unify the "operation cost" and "risk avoidance" into the speed selection process.

[0100] S500, based on the optimal avoidance speed vector, the ship is navigated to avoid the risk.

[0101] Specifically, as the last link of the process of the present application, the goal is to convert the optimal avoidance speed vector output by the previous step (S400) into a standardized form into an executable avoidance instruction, and issue it through the ship control interface, while establishing a real-time feedback mechanism to enable the system to operate in a closed loop.

[0102] Input: optimal avoidance speed vector , representing the safest and least operation cost combination of speed and heading in the current frame; minimum cost value , reflecting the comprehensive cost of the speed instruction in terms of collision avoidance efficiency, energy consumption and operation difficulty; in addition, it also relies on the ship current state vector generated by S100, for generating the difference feedback of the expected control amount.

[0103] Further, this step first converts into a standard steering instruction format, which is suitable for different types of navigation control system interfaces. This conversion is based on the thrust-rudder mapping function provided by the shipyard, which is a set of calibration lookup table data, with the target speed and heading as input, and the corresponding propulsion power setting value and rudder angle target The conversion calculation does not need to be modeled online, but is directly solved by interpolation from a static mapping table, as follows:

[0104] ;

[0105] where is the control instruction to be executed in the current frame, which consists of two quantities: the propulsion power and the rudder angle set value . This control quantity is packaged in NMEA or IEC61162-450 encoding format and sent to the ship's electronic navigation control module or intelligent course keeping system through the CAN bus. This control interface has been deployed in medium and large ice area ships and has standard driving capability, without the need for user manual intervention.

[0106] The system will then read the actual state after execution at time , which is output by the same GNSS / IMU fusion module in S100 to ensure alignment with the instruction execution time. The execution error is calculated based on this:

[0107] ;

[0108] This error is used to determine whether the resolution of the next round of feasible speed search needs to be adjusted, or the parameter value is dynamically increased to enhance the model's fault tolerance to propulsion hysteresis. For example, if knor continues for more than two frames, the system will automatically increase the grid precision of the feasible speed domain from the original kn× to kn× , while increasing by 5% to enhance the subsequent adaptability to hysteresis dynamics.

[0109] The execution error is also written into the task log as an indicator of the system's "controllability margin" for use by the captain or for system analysis by the background.

[0110] As shown in Figure 3 , the CRI, speed, relative distance, DCPA, and TCPA dynamic curves in a typical polar two-ship encounter process are shown. As can be seen from the figure, each indicator is always within the controllable range, the risk index does not exceed the warning threshold, and the relative distance maintains a safe margin, indicating that the real-time control effect of this scheme on the intersection of two ships is good, and the navigation process is safe and reliable.

[0111] It should be noted that the above-described workflow is merely illustrative and does not limit the scope of protection of the present application. In actual applications, a person skilled in the art can select part or all of the above-described workflow to achieve the purpose of the present embodiment according to actual needs, which is not limited herein.

[0112] In addition, technical details not described in detail in the present embodiment can be found in the parameter operation method provided by any embodiment of the present application, which will not be described here.

[0113] It should be noted that in this document, the terms "comprise", "contain" or any other variant thereof are intended to cover non-exclusive inclusion, so that the process, method, article or system including a series of elements not only includes those elements, but also includes other elements not explicitly listed or inherent to such process, method, article or system. Without more limitations, the element defined by the statement "comprises a" does not exclude the presence of another identical element in the process, method, article or system that includes the element.

[0114] The above-mentioned embodiment numbers of the present application are only for description, not representing the advantages or disadvantages of the embodiments.

[0115] From the above description of the embodiments, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be realized by means of software and necessary general hardware platform, of course, they can also be realized by hardware, but in many cases the former is a better embodiment. Based on such understanding, the technical solutions of the present application or the part that contributes to the prior art can be embodied in the form of software product, which is stored in a storage medium (such as read-only memory / random access memory, magnetic disk, optical disk), including a plurality of instructions to make a terminal device (which can be a mobile phone, computer, server, air conditioner, or network device, etc.) execute the method described in each embodiment of the present application.

[0116] The above is only the preferred embodiment of the present application, and does not limit the patent scope of the present application. Any equivalent structure or equivalent process transformation, or direct or indirect application in other related technical fields, is also included in the patent protection scope of the present application.

Claims

1. A method for quantifying the risk of ship navigation under the three-dimensional time-delay speed barrier of polar sea ice, characterized in that, The method comprises: S100, collecting ship local sensor data to form a ship motion state and an ice block motion set; S200, constructing a three-dimensional-time delay velocity obstacle model TDIVO³D processing the ship motion state and the ice block motion set, generating two three-dimensional binary mask maps, including a velocity obstacle domain and a feasible velocity domain, specifically: Obtaining the maximum propulsion rate and the maximum turning angle rate of the ship, and constructing a discretized two-dimensional velocity-course grid; Using Euler forward integration for each candidate velocity vector in the two-dimensional velocity-course grid to perform state deduction in a 60-second time window, obtaining a predicted spatial path of the ship within the next 60S, and introducing a delay correction amount to delay all paths by the time of the delay correction amount, obtaining a feasible velocity domain after delay compensation; According to the hull safety domain in the horizontal direction and the ice block circumscribed radius, a comprehensive safety radius of each ice block is constructed; Performing parallel Minkowski summation calculation on each candidate velocity vector on the GPU to obtain the velocity obstacle domain; S300, mapping the velocity obstacle domain and the feasible velocity domain into a continuous quantitative index sensitive to polar navigation risk; S400, designing a target function to minimize collision avoidance risk or angle change according to the velocity obstacle domain and the feasible velocity domain and the continuous quantitative index, calculating the target function, and selecting the velocity vector corresponding to the minimum value as the optimal avoidance velocity vector; S500, based on the optimal avoidance velocity vector, the ship performs navigation risk avoidance.

2. The method of quantifying the risk of ship navigation under the three-dimensional-time-delay speed obstacle of polar sea ice according to claim 1, characterized in that, The ship local sensor data includes a sequence of space-borne synthetic aperture radar images, a rough ship attitude provided by GNSS / IMU combined navigation, a speed and course information output by a ship-borne AIS system, and a green laser composite point cloud of a multi-beam sonar in front of the ship's bow, and the ship local sensor data is uniformly mapped to a local coordinate system with the ship's center of mass as the origin.

3. The method of quantifying the navigation risk of a ship under the three-dimensional-time-delay speed barrier of polar sea ice according to claim 2, characterized in that, The ship local sensor data is collected to form a ship motion state and an ice block motion set, specifically: Receiving ship local sensor data according to a preset ship local data fusion module; Pretreating the ship local sensor data to form a ship motion state and an ice block motion set; wherein the pretreatment includes coherent difference processing, iterative closest point rigid body matching, and extended Kalman filter processing; The ship motion state includes a fusion state vector of the current frame of the ship and a ship center of mass path; the ice block motion set includes a fusion state vector of the current frame of the ice block and an ice block center of mass path.

4. The method of quantifying the navigation risk of a ship under the three-dimensional-time-delay speed barrier of polar sea ice according to claim 1, characterized in that, The parallel Minkowski summation calculation on each candidate velocity vector on the GPU to obtain the velocity obstacle domain is specifically: For each ice, determine if the predicted path of the vessel will enter the corresponding integrated safety radius of the ice within the sliding time window If the vessel is moving at a speed less than the integrated safety radius, mark the speed vector as a dangerous point and include it in the speed obstacle region.

5. The method of quantifying the risk of ship navigation under the three-dimensional-time-delay speed obstacle of polar sea ice according to claim 1, characterized in that, The velocity obstacle domain and the feasible velocity domain are mapped into a continuous quantitative index sensitive to polar navigation risk, specifically: The sum operation of the two mask voxel sets of the velocity obstacle domain and the feasible velocity domain is completed in parallel on the GPU to obtain a volume ratio; in an ideal navigation state, the volume ratio is less than 0.1; once the volume ratio exceeds 0.3, it indicates that there is no safe feasible solution in the feasible velocity domain for the ship. When the navigation state has a volume ratio of less than 0.1, the positioning covariance of the velocity obstacle domain is... Convert to scalar That is, take the positioning covariance. traces As an overall error index, the continuous quantization index is calculated using the following exponential function. : ； wherein, the steepness of the growth of the risk, is the decision risk benchmark, i.e. times the uncertainty, after which the risk starts to rise significantly; is the mapping scale, used to normalize the covariance dimension to a dimensionless volume proportion matching interval.

6. The method of quantifying the navigation risk of a ship under the three-dimensional-time-delay speed barrier of polar sea ice according to claim 5, characterized in that, The exponential function is used to ensure that when the mapping scale is small, even a certain degree of covariance will not amplify the risk to an unusable state; but when the mapping scale is close to and the covariance increases, the continuous quantification index will quickly approach 1, reminding the system or the driver that the current speed domain is almost unable to meet the requirement for safe travel.

7. The method of quantifying the risk of ship navigation under the three-dimensional-time-delay speed obstacle of polar sea ice according to claim 1, characterized in that, The target function is designed to minimize collision risk or angle change according to the speed obstacle region and the feasible speed region and a continuous quantization index, the target function is calculated, and a speed vector corresponding to a minimum value is selected as an optimal avoidance speed vector, and the specific process is as follows: A speed safety sub-region is obtained by eliminating the speed obstacle region from the feasible speed region, wherein the speed safety sub-region includes a plurality of candidate speed vectors, and the candidate speed vectors include speed and angle. According to the candidate velocity vector and the continuous quantization index A target function is designed to minimize the risk of collision avoidance or angle change, and a response hysteresis cost function is introduced into the target function, representing the required response time of the current velocity vector. The optimal avoidance velocity vector and the corresponding minimum generation value are calculated through the target function. The response hysteresis cost function is a dynamic function, and in the polar low-temperature environment, the response time of low-temperature propulsion, the rudder angle response time, the current maximum acceleration and maximum turning ability of the ship and the candidate speed vector are used to calculate the response hysteresis cost function.

8. The method of quantifying the risk of ship navigation under the three-dimensional-time-delay speed obstacle of polar sea ice according to claim 1, characterized in that, The ship is navigated to avoid risks based on the optimal avoidance speed vector, and the specific process is as follows: The optimal avoidance speed vector is converted into a standard steering instruction format; According to the standard steering instruction format and the preset calibration lookup table data, the corresponding propulsion power set value and rudder angle target are obtained by interpolation calculation; executing the propulsion power set value and the rudder angle target, and reading back the actual ship state at the moment when the execution is completed; The execution error is calculated based on the state of the ship and the optimal avoidance speed vector; The resolution of the feasible speed search is adjusted based on the execution error.

9. The method of quantifying the navigation risk of a ship under the three-dimensional-time-delay speed barrier of polar sea ice according to claim 8, characterized in that, The standard steering instruction format is packaged in NMEA or IEC61162-450 encoding format and sent to the ship's electronic control navigation control module or intelligent course keeping system through the CAN bus.

Images