A Smart Collision Avoidance Decision-Making Method and System for Ships

Abstract

This invention discloses an intelligent collision avoidance decision-making method and system for ships, relating to the field of intelligent ship navigation technology. The method includes: collecting radar data and AIS information of the ship; constructing a dynamic prediction model based on the radar data and AIS information; predicting the future trajectory of a target ship using a Kalman filter algorithm; calculating the encounter parameters between the two ships by combining the ship's own navigation parameters; establishing a risk assessment matrix based on the encounter parameters; and using a fuzzy neural network algorithm to determine the risk level of the current navigation situation; based on the risk level determination result, calculating collision avoidance maneuver parameters using an adaptive particle swarm optimization algorithm; and outputting a collision avoidance action plan. This invention achieves full automation from information perception, situation prediction, risk assessment to decision execution, significantly reducing the workload of maritime personnel and improving navigation safety.

Description

Technical Field

[0001] This invention relates to the field of intelligent ship navigation technology, and in particular to an intelligent collision avoidance decision-making method and system for ships. Background Technology

[0002] Traditional ship collision avoidance decision-making relies heavily on crew experience and judgment, inherently suffering from high subjectivity and low efficiency. In recent years, with the continuous upgrading of shipborne navigation equipment, intelligent collision avoidance technology based on radar data and AIS information has gradually developed. Current research on collision avoidance decision-making mainly focuses on rule-based expert systems, deep learning-based situational awareness, and dynamic programming-based path planning. Among these, intelligent collision avoidance systems integrating multi-source data have shown significant advantages in improving decision reliability, but they still face many challenges when dealing with multi-ship encounters in complex navigation environments.

[0003] Existing ship collision avoidance decision-making methods generally suffer from the following technical problems: First, the dynamic changes in data reliability are not fully considered during multi-source data fusion, leading to unstable situational awareness accuracy. Second, ship motion prediction models do not adequately consider the impact of environmental factors such as wind, waves, and currents, affecting the accuracy of predicted trajectories. Third, existing risk assessment methods often use simple distance threshold judgments, lacking in-depth analysis of ship motion trends, making it difficult to accurately reflect the actual collision avoidance risk level. Finally, the collision avoidance decision optimization process fails to effectively balance the relationship between safety and economy, resulting in problems of over-avoidance or under-avoidance in collision avoidance operations.

[0004] To address the aforementioned problems, this invention proposes an intelligent collision avoidance decision-making method and system for ships. This method aims to achieve more accurate and reliable intelligent collision avoidance decisions by constructing an adaptive data fusion model, introducing a motion prediction method that corrects for environmental factors, designing a multi-level risk assessment mechanism, and optimizing the collision avoidance parameter calculation strategy. Summary of the Invention

[0005] In view of the problems existing in ship collision avoidance decision-making methods, this invention is proposed.

[0006] Therefore, the problem to be solved by this invention is how to achieve high-precision and high-reliability intelligent collision avoidance decision-making for ships, and effectively deal with multi-ship encounters in complex navigation environments.

[0007] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0008] In a first aspect, embodiments of the present invention provide an intelligent collision avoidance decision-making method for ships, comprising: collecting radar data and AIS information of the ship; constructing a ship dynamic prediction model based on the radar data and AIS information, predicting the future navigation trajectory of the target ship using a Kalman filter algorithm, and calculating the encounter parameters of the two ships in conjunction with the ship's own navigation parameters; establishing a risk assessment matrix based on the encounter parameters, and using a fuzzy neural network algorithm to determine the risk level of the current navigation situation; and calculating collision avoidance maneuver parameters using an adaptive particle swarm optimization algorithm based on the risk level determination result, and outputting a collision avoidance action plan.

[0009] As a preferred embodiment of the intelligent collision avoidance decision-making method for ships described in this invention, the method includes: calculating collision avoidance maneuvering parameters based on risk level assessment results and outputting a collision avoidance action plan using an adaptive particle swarm optimization algorithm, comprising: establishing a collision avoidance constraint condition library based on risk level assessment results and simultaneously establishing a collision avoidance parameter calculation model using the adaptive particle swarm optimization algorithm, wherein the collision avoidance constraint condition library includes avoidance rules and ship maneuvering performance; inputting encounter parameters and the future navigation trajectory of the target ship into the collision avoidance parameter calculation model, verifying the combination of collision avoidance maneuvering parameters, and outputting the turning angle value and speed adjustment amount corresponding to the combination of collision avoidance maneuvering parameters to form a collision avoidance action plan.

[0010] As a preferred embodiment of the intelligent collision avoidance decision-making method for ships described in this invention, the collision avoidance maneuver parameters include the optimal turning angle and speed adjustment; the collision avoidance parameter calculation model is constructed by using the collision avoidance constraint library as the optimization objective function to construct a search space including the turning angle and speed adjustment, and setting the initial position distribution and initial velocity distribution of the particle swarm; using adaptive weight coefficients to update the position and velocity of the particle swarm, and calculating the collision avoidance trajectory corresponding to the particle swarm; based on the collision avoidance trajectory, using collision avoidance safety margin, route deviation, and maneuvering energy consumption as evaluation indicators, outputting the individual optimal position and global optimal position of the particle swarm; and iteratively calculating the particle swarm according to the evaluation indicators to obtain a collision avoidance parameter combination that satisfies the avoidance rules.

[0011] As a preferred embodiment of the intelligent collision avoidance decision-making method for ships described in this invention, the method includes: establishing a risk assessment matrix based on the encounter parameters, and using a fuzzy neural network algorithm to determine the risk level of the current navigation situation, including: constructing a risk assessment matrix and dividing the risk assessment matrix into level standards according to the threshold range of the encounter parameters, wherein the encounter parameters include encounter time and encounter distance; constructing a risk assessment model based on the risk assessment matrix using a fuzzy neural network, wherein the fuzzy neural network includes an input layer, a fuzzy layer, a rule layer, and an output layer; training the risk assessment model based on historical collision avoidance case data, and optimizing the model weight parameters and adjusting the membership function through a backpropagation algorithm; inputting the current navigation situation data into the trained risk assessment model, calculating the membership value of the collision risk, and outputting the risk level of the current situation according to the principle of maximum membership; and continuously collecting navigation situation data based on the risk level of the current situation, and adjusting and updating the risk level determination results in real time.

[0012] As a preferred embodiment of the intelligent collision avoidance decision-making method for ships described in this invention, the risk assessment matrix is ​​divided as follows: when the encounter time is less than a first time threshold and the encounter distance is less than a first distance threshold, the risk assessment matrix corresponds to an emergency danger level; when the encounter time is less than a second time threshold and the encounter distance is less than a second distance threshold, the risk assessment matrix corresponds to a high danger level; when the encounter time is between a second time threshold and a third time threshold and the encounter distance is between a second distance threshold and a third distance threshold, the risk assessment matrix corresponds to a medium danger level; when the encounter time is greater than a third time threshold or the encounter distance is greater than a third distance threshold, the risk assessment matrix corresponds to a low danger level; and when the encounter time is greater than a fourth time threshold and the encounter distance is greater than a fourth distance threshold, the risk assessment matrix corresponds to a safe level.

[0013] As a preferred embodiment of the intelligent collision avoidance decision-making method for ships described in this invention, the method includes: constructing a ship dynamic prediction model based on radar data and AIS information; predicting the future trajectory of the target ship using a Kalman filter algorithm; and calculating the encounter parameters of the two ships by combining the ship's own navigation parameters. This includes: inputting radar data and AIS information into the ship dynamic prediction model and establishing a state vector including the ship's longitudinal velocity, lateral velocity, and bow angular velocity, wherein the ship dynamic prediction model uses a six-degree-of-freedom equation of motion to describe the ship's navigation state; constructing a state transition matrix for the Kalman filter based on the state vector; and simultaneously establishing an observation equation to incorporate radar data and AIS information. The measured values ​​of the information are mapped to the state space; the state of the target ship is iteratively updated using the Kalman filter, and the state estimate is optimized through a prediction-correction loop to output the future navigation trajectory of the target ship; based on the future navigation trajectory of the target ship, a relative motion model is established in conjunction with the navigation parameters of the ship itself, the predicted trajectory of the target ship is transformed into a relative motion coordinate system, and the relative motion trajectory curve is fitted using the least squares method; the meeting parameters of the two ships are calculated and output according to the relative motion trajectory curve; uncertainty analysis is performed on the future navigation trajectory of the target ship and the meeting parameters, and the corresponding confidence interval is output according to the error propagation characteristics of the state estimate.

[0014] As a preferred embodiment of the intelligent collision avoidance decision-making method for ships described in this invention, the radar data includes the echo signal and distance information of the target ship; the AIS information includes the latitude and longitude positions, speed values ​​and heading angles of the ship and the target ship.

[0015] Secondly, embodiments of the present invention provide an intelligent collision avoidance decision-making system for ships, comprising: a data acquisition module for acquiring radar data and AIS information of the ship; a target ship trajectory prediction module for constructing a dynamic prediction model of the ship based on the radar data and AIS information, predicting the future navigation trajectory of the target ship using a Kalman filter algorithm, and calculating the encounter parameters of the two ships in conjunction with the ship's navigation parameters; a judgment module for establishing a risk assessment matrix based on the encounter parameters, and using a fuzzy neural network algorithm to judge the risk level of the current navigation situation; and a navigation risk assessment module for calculating collision avoidance maneuver parameters using an adaptive particle swarm optimization algorithm based on the risk level judgment result, and outputting a collision avoidance action plan.

[0016] Thirdly, embodiments of the present invention provide a computer device, including a memory and a processor, wherein the memory stores a computer program, and the computer program instructions, when executed by the processor, implement the steps of the intelligent collision avoidance decision-making method for ships as described in the first aspect of the present invention.

[0017] Fourthly, embodiments of the present invention provide a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program instructions are executed by a processor, they implement the steps of the intelligent collision avoidance decision-making method for ships as described in the first aspect of the present invention.

[0018] The beneficial effects of this invention are as follows: It constructs a comprehensive and reliable information foundation by fusing radar and AIS information from multiple sources; it utilizes the Kalman filter algorithm to build a dynamic prediction model, achieving high-precision prediction of the future trajectory of target vessels; it establishes a risk assessment matrix based on encounter parameters and combines it with a fuzzy neural network algorithm to intelligently determine the risk level of the current navigation situation; it employs an adaptive particle swarm optimization algorithm to calculate collision avoidance maneuver parameters, generating an optimal collision avoidance scheme that satisfies international maritime collision avoidance rules while also considering vessel maneuverability; this invention achieves full-process automation from information perception, situation prediction, risk assessment to decision execution, greatly reducing the workload of maritime personnel and improving navigation safety. Attached Figure Description

[0019] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. Wherein:

[0020] Figure 1 This is the overall flowchart of the intelligent collision avoidance decision-making method for ships in Example 1. Detailed Implementation

[0021] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0022] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0023] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0024] Example 1

[0025] Reference Figure 1This is the first embodiment of the present invention, which provides an intelligent collision avoidance decision-making method for ships, including,

[0026] S1: Collect radar data and AIS information for this vessel.

[0027] Specifically, radar data includes the echo signal and range information of the target vessel. Adaptive CFAR processing is applied to the radar data to suppress the influence of sea clutter and extract valid target echo information.

[0028] It should be noted that the shipborne X-band radar scanning system continuously transmits electromagnetic wave signals, records the time interval of the reflected echoes and converts them into distance data of the target ship, and simultaneously measures and digitizes the intensity of the echo signals; the radial velocity of the target ship is calculated using the Doppler effect principle, and the relative azimuth angle data of the target ship is obtained by combining the radar scanning azimuth.

[0029] Furthermore, AIS information includes the latitude and longitude positions, speed values, and heading angles of both the vessel and the target vessel.

[0030] It should be noted that the system receives the latitude and longitude coordinates output by the ship's GPS positioning system, obtains heading data through a gyrocompass, collects the ship's ground speed value using a Doppler log, and integrates the positioning data, heading data, and speed data into the ship's status information; it also monitors the AIS communication channel, receives dynamic information messages broadcast by the target ship at specified time intervals, parses the messages to obtain the target ship's latitude and longitude position, speed value, and heading angle, and establishes a dynamic information database of the target ship;

[0031] Furthermore, the collected radar data and AIS information are synchronized in time and have their coordinates unified, outliers are removed and missing data is filled in.

[0032] S2: Construct a ship dynamic prediction model based on radar data and AIS information, predict the future navigation trajectory of the target ship using the Kalman filter algorithm, and calculate the encounter parameters between the two ships by combining the navigation parameters of the ship itself.

[0033] Specifically, radar data and AIS information are input into the ship dynamic prediction model, and a state vector including the ship's longitudinal speed, lateral speed and bow angular velocity is established. The ship dynamic prediction model uses a six-degree-of-freedom motion equation to describe the ship's navigation state.

[0034] It should be noted that the meeting parameters include the meeting time and the meeting distance.

[0035] The preferred formula for the ship dynamic prediction model is as follows:

[0036]

[0037] in, The coordinates of the ship in the east direction, The coordinates of the ship in the north direction. Let u be the ship's bow angle, u be the ship's longitudinal velocity, v be the ship's transverse velocity, r be the ship's angular velocity of turning, and V be the ship's bow angle. w For wind speed, θ w θ is the wind direction angle, β is the correction factor, and θ is the wind direction angle. c Indicates the direction of the ocean current.

[0038] Furthermore, a state transition matrix for the Kalman filter is constructed based on the state vector, and an observation equation is established to map the measured values ​​of radar data and AIS information to the state space.

[0039] Preferably, the specific formula for the observation equation is as follows:

[0040]

[0041] Among them, z k Let x be the observation vector. k Let k be the eastward coordinate, y k Let ψ be the northward coordinate at time k. k Let be the heading angle at time k, α1 be the error coefficient for eastward position measurement, α2 be the error coefficient for northward position measurement, α3 be the error coefficient for heading angle measurement, and R be the error coefficient for eastward position measurement. k w is the target distance k To observe noise.

[0042] Furthermore, the Kalman filter is used to iteratively update the state of the target ship. The state estimate is optimized through a prediction-correction loop to output the future navigation trajectory of the target ship. Based on the future navigation trajectory of the target ship, a relative motion model is established in combination with the ship's own navigation parameters. The predicted trajectory of the target ship is transformed into a relative motion coordinate system, and the least squares method is used to fit the relative motion trajectory curve.

[0043] It should be noted that the relative motion model is based on...

[0044] Specifically, the meeting parameters of the two ships are calculated and output based on the relative motion trajectory curves; uncertainty analysis is performed on the future navigation trajectory and meeting parameters of the target ship, and the corresponding confidence interval is output based on the error propagation characteristics of the state estimate.

[0045] S3: Based on the encounter parameters, establish a risk assessment matrix and use a fuzzy neural network algorithm to determine the risk level of the current navigation situation;

[0046] Specifically, a risk assessment matrix is ​​constructed, and the risk assessment matrix is ​​graded according to the threshold range of the encounter parameters.

[0047] Preferably, the risk assessment matrix is ​​divided as follows: when the meeting time is less than a first time threshold and the meeting distance is less than a first distance threshold, it corresponds to an emergency risk level in the risk assessment matrix; when the meeting time is less than a second time threshold and the meeting distance is less than a second distance threshold, it corresponds to a high risk level in the risk assessment matrix; when the meeting time is between a second time threshold and a third time threshold and the meeting distance is between a second distance threshold and a third distance threshold, it corresponds to a medium risk level in the risk assessment matrix; when the meeting time is greater than a third time threshold or the meeting distance is greater than a third distance threshold, it corresponds to a low risk level in the risk assessment matrix; and when the meeting time is greater than a fourth time threshold and the meeting distance is greater than a fourth distance threshold, it corresponds to a safe level in the risk assessment matrix.

[0048] It should be noted that if the membership value of the risk level corresponds to the emergency danger level, the emergency collision avoidance plan will be activated and an alarm signal will be issued to the driver; if the membership value of the risk level corresponds to the high danger level, a collision avoidance suggestion will be generated and a collision avoidance countdown will be displayed to the driver; if the membership value of the risk level corresponds to the medium danger level, the encounter parameters will be continuously monitored and a collision avoidance plan will be prepared; if the membership value of the risk level corresponds to the low danger level or the safe level, navigation status monitoring will be maintained.

[0049] Furthermore, based on the risk assessment matrix, a risk assessment model is constructed using a fuzzy neural network, which includes an input layer, a fuzzy layer, a rule layer, and an output layer.

[0050] It should be noted that the fuzzy layer sets the membership function of the encounter parameter; the rule layer embeds an expert knowledge rule base.

[0051] Furthermore, a risk assessment model is trained based on historical collision avoidance case data, and the model weight parameters are optimized and the membership function is adjusted through backpropagation algorithm. The current navigation situation data is input into the trained risk assessment model to calculate the membership value of the collision risk, and the risk level of the current situation is output according to the principle of maximum membership.

[0052] The preferred formula for the membership degree value is as follows:

[0053]

[0054] Where μ is the risk membership value, γ1 is the DCPA influence weight coefficient, DCPA is the nearest encounter distance, and DCPA is the distance to the nearest meeting point. s Where γ is the safety threshold distance, γ2 is the speed influence weighting coefficient, and V r V is the relative velocity. c This is the critical safe speed.

[0055] Specifically, based on the current risk level, navigation situation data will be continuously collected, and the risk level assessment results will be adjusted and updated in real time.

[0056] S4: Based on the risk level assessment results, use the adaptive particle swarm optimization algorithm to calculate the collision avoidance maneuver parameters and output the collision avoidance action plan.

[0057] Specifically, a collision avoidance constraint library is established based on the risk level assessment results, and an adaptive particle swarm optimization algorithm is used to establish a collision avoidance parameter calculation model. The collision avoidance constraint library includes avoidance rules and ship maneuvering performance.

[0058] Furthermore, the collision avoidance parameter calculation model is constructed as follows: using a collision avoidance constraint library as the optimization objective function, a search space including steering angle and speed adjustment is built, and the initial position and initial velocity distribution of the particle swarm are set; adaptive weight coefficients are used to update the position and velocity of the particle swarm, and the corresponding collision avoidance trajectory is calculated; based on the collision avoidance trajectory, the collision avoidance safety margin, the route deviation, and the control energy consumption are used as evaluation indicators to output the individual optimal position and the global optimal position of the particle swarm; the particle swarm is iteratively calculated according to the evaluation indicators to obtain the collision avoidance parameter combination that satisfies the avoidance rules.

[0059] Furthermore, the collision avoidance parameter calculation model inputs the encounter parameters and the future trajectory of the target ship, verifies the collision avoidance maneuver parameter combination, and outputs the corresponding turning angle value and speed adjustment amount to form a collision avoidance action plan.

[0060] It should be noted that the collision avoidance maneuver parameters include the optimal turning angle and the speed adjustment.

[0061] Specifically, the collision avoidance maneuver parameters are verified, including: when the turning angle exceeds the ship's maximum turning angle, the turning angle is limited to the range of the ship's maximum turning angle; when the speed adjustment exceeds the ship's maximum deceleration or acceleration range, the speed adjustment is limited to the range allowed by the ship's maneuverability; when the minimum distance between the calculated collision avoidance trajectory and the predicted trajectory of the target ship is less than the safe distance, the collision avoidance parameters are recalculated for optimization; when the collision avoidance trajectory deviates too much from the original planned route, a collision avoidance parameter combination with a smaller deviation is selected under the premise of ensuring safety; when the collision avoidance maneuver parameters meet the ship's maneuverability constraints and can ensure collision avoidance safety, the corresponding turning angle value and speed adjustment are output as the final collision avoidance action plan.

[0062] Furthermore, this embodiment also provides an intelligent collision avoidance decision-making system for ships, including: a data acquisition module for acquiring radar data and AIS information of the ship; a target ship trajectory prediction module for constructing a ship dynamic prediction model based on the radar data and AIS information, predicting the future navigation trajectory of the target ship using a Kalman filter algorithm, and calculating the encounter parameters of the two ships in combination with the ship's navigation parameters; a judgment module for establishing a risk assessment matrix based on the encounter parameters, and using a fuzzy neural network algorithm to judge the risk level of the current navigation situation; and a navigation risk assessment module for calculating collision avoidance maneuver parameters using an adaptive particle swarm optimization algorithm based on the risk level judgment result, and outputting a collision avoidance action plan.

[0063] This embodiment also provides a computer device applicable to intelligent collision avoidance decision-making methods for ships, including a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to realize the intelligent collision avoidance decision-making method for ships as proposed in the above embodiment.

[0064] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0065] In summary, this invention constructs a comprehensive and reliable information foundation by fusing radar and AIS information from multiple sources; it utilizes the Kalman filter algorithm to build a dynamic prediction model, achieving high-precision prediction of the future trajectory of target vessels; it establishes a risk assessment matrix based on encounter parameters and combines it with a fuzzy neural network algorithm to intelligently determine the risk level of the current navigation situation; it employs an adaptive particle swarm optimization algorithm to calculate collision avoidance maneuver parameters, generating an optimal collision avoidance scheme that satisfies international maritime collision avoidance rules while also considering vessel maneuverability; this invention achieves full-process automation from information perception, situation prediction, risk assessment to decision execution, greatly reducing the workload of maritime personnel and improving navigation safety.

[0066] Example 2

[0067] Referring to Table 1, the second embodiment of the present invention provides an intelligent collision avoidance decision-making method for ships. To verify the beneficial effects of the present invention, scientific demonstration is carried out through economic benefit calculation and simulation experiments.

[0068] Specifically, this experiment selected a 5,000-ton container ship as the subject and set up multiple target ship scenarios in the sea area. The experimental equipment included an X-band radar system (model: XR-420, detection range 24 nautical miles), an AIS receiver (model: AIS-M7, reception range 30 nautical miles), and a self-developed ship collision avoidance decision system (ICADS-V2.0). The ship collision avoidance decision system was initialized and configured, including setting the parameters of the radar and AIS data acquisition modules, adjusting the parameters of the Kalman filter, determining the threshold of the risk assessment matrix, and optimizing the parameters of the particle swarm algorithm.

[0069] Furthermore, in the data acquisition and processing phase, the experimental system acquired radar data and AIS information at a frequency of 2Hz. Radar data included the target vessel's echo signal and range information, while AIS information included the vessel's latitude and longitude, speed, and heading angle. An adaptive noise reduction algorithm was used to process the radar echo signal, and the AIS data was checked for integrity, eliminating outliers. In the dynamic prediction model construction phase, a dynamic prediction model for the vessel was built based on six-degree-of-freedom motion equations. This model considered state variables such as the vessel's longitudinal speed, lateral speed, and bow angular velocity. The state transition matrix of the Kalman filter was designed as a time-varying matrix to adapt to the dynamic changes in the navigation environment. The observation noise covariance matrix was determined through multiple experimental optimizations. Through a prediction-correction loop, the system iteratively updated the target vessel's state, with a prediction time window set at 10 minutes and a sampling interval of 30 seconds. In the risk assessment phase, a risk assessment matrix was constructed based on encounter parameters, and five risk levels were set: critical danger, high danger, moderate danger, low danger, and safe. The threshold ranges for the risk assessment matrix are as follows: first time threshold 3 minutes, first distance threshold 0.5 nautical miles; second time threshold 6 minutes, second distance threshold 1 nautical mile; third time threshold 12 minutes, third distance threshold 2 nautical miles; fourth time threshold 20 minutes, fourth distance threshold 3 nautical miles. The fuzzy neural network used in the experiment has a four-layer structure, containing 10 input neurons, 25 fuzzy neurons, 30 regular neurons, and 5 output neurons. The model was trained using 500 sets of historical collision avoidance case data, with a learning rate of 0.05 and 1000 training iterations. The final model classification accuracy reached 94.3%.

[0070] Furthermore, in the collision avoidance decision generation stage, the experimental system established a collision avoidance constraint library based on risk levels, including rules for cross-encounter, head-on, and overtaking in the International Regulations for Preventing Collisions at Sea (ICP-35), as well as the vessel's maneuvering performance parameters. The adaptive particle swarm optimization algorithm was configured with the following parameters: 50 particles, a maximum of 100 iterations, and an inertia weight range of [0.4, 0.9]. A search space was constructed, including turning angles [-60°, 60°] and speed adjustments [-5kn, +2kn]. Collision safety margin (weight 0.5), course deviation (weight 0.3), and maneuvering energy consumption (weight 0.2) were used as evaluation indicators. Through adaptive weight adjustment and iterative optimization, the optimal collision avoidance action plan that meets the collision avoidance requirements was finally output.

[0071] Specifically, as shown in Table 1, the analysis of experimental data demonstrates significant technical advantages of the method of this invention. In terms of decision-making timeliness, the average collision avoidance decision time is 13.87 seconds, an improvement of approximately 65%, and it can still maintain high decision-making efficiency in complex multi-ship encounter scenarios. In terms of collision avoidance safety, the minimum encounter distance in all scenarios is greater than the safe distance (0.5 nautical miles) stipulated by the COLREG Convention, with an average minimum encounter distance of 0.95 nautical miles, which fully guarantees navigation safety.

[0072] Table 1. Experimental Data Table

[0073] Encounter Types Initial encounter distance relative speed Collision avoidance time Minimum meeting distance route deviation Decision accuracy Cross Encounter 2.5 nautical miles 15 sections 12.3s 1.2 nautical miles 5.8％ 96.5％ Chasing and encountering 1.8 nautical miles 8 sections 15.6s 0.8 nautical miles 4.2％ 97.8％ Encounter 2.2 nautical miles 12 sections 13.8s 1.1 nautical miles 6.1％ 95.9％ Cross 1.5 nautical miles 18 sections 10.2s 0.9 nautical miles 7.3％ 98.2％ Oblique meeting 2.0 nautical miles 14 sections 14.5s 1.0 nautical miles 5.5％ 96.8％ Multiple ships meet 1.7 nautical miles 16 sections 16.8s 0.7 nautical miles 8.2％ 94.5％

[0074] Furthermore, regarding route deviation, this method optimizes collision avoidance parameters through an adaptive particle swarm optimization algorithm, resulting in an average route deviation of only 6.18%, a reduction of approximately 65%. Particularly in overtaking encounter scenarios, a safe avoidance requires only a 4.2% route deviation, significantly improving navigation efficiency. In terms of decision accuracy, the average accuracy for six typical encounter scenarios reaches 96.62%, and even in the most complex multi-ship encounter scenario, it maintains a high accuracy of 94.5%, demonstrating the robustness and reliability of this method in complex navigation environments.

[0075] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A ship intelligent collision avoidance decision-making method, characterized in that: include, Collect radar data and AIS information of this vessel; Based on the radar data and AIS information, a ship dynamic prediction model is constructed. The future navigation trajectory of the target ship is predicted by the Kalman filter algorithm, and the encounter parameters of the two ships are calculated by combining the navigation parameters of the ship itself. Based on the encounter parameters, a risk assessment matrix is ​​established, and a fuzzy neural network algorithm is used to determine the risk level of the current navigation situation. Based on the risk level assessment results, the collision avoidance maneuver parameters are calculated using an adaptive particle swarm optimization algorithm, and a collision avoidance action plan is output. A collision avoidance constraint library is established based on the risk level assessment results, and a collision avoidance parameter calculation model is established using an adaptive particle swarm optimization algorithm. The collision avoidance constraint library includes avoidance rules and ship maneuvering performance. The collision avoidance parameter calculation model is input into the encounter parameters and the future navigation trajectory of the target ship. The collision avoidance maneuver parameter combination is verified, and the turning angle value and speed adjustment amount corresponding to the collision avoidance maneuver parameter combination are output to form a collision avoidance action plan. The collision avoidance maneuver parameters include the optimal turning angle and speed adjustment; the method for constructing the collision avoidance parameter calculation model is as follows: Using the aforementioned collision avoidance constraint library as the optimization objective function, a search space including steering angle and speed adjustment is constructed, and the initial position distribution and initial velocity distribution of the particle swarm are set. Adaptive weighting coefficients are used to update the position and velocity of the particle swarm, and the collision avoidance trajectory corresponding to the particle swarm is calculated. Based on the collision avoidance trajectory, the collision avoidance safety margin, the deviation from the flight path, and the control energy consumption are used as evaluation indicators to output the individual optimal position and the global optimal position of the particle swarm. The particle swarm is iteratively calculated based on the evaluation index to obtain a combination of collision avoidance parameters that satisfy the avoidance rules. Radar data and AIS information are input into the ship dynamic prediction model, and a state vector including the ship's longitudinal velocity, lateral velocity and bow angular velocity is established. The ship dynamic prediction model uses a six-degree-of-freedom equation of motion to describe the ship's navigation state. Based on the state vector, a state transition matrix of the Kalman filter is constructed, and an observation equation is established to map the measured values ​​of radar data and AIS information to the state space. The Kalman filter is used to iteratively update the state of the target ship, and the state estimate is optimized by prediction-correction loop to output the future navigation trajectory of the target ship. Based on the future navigation trajectory of the target vessel, a relative motion model is established by combining the navigation parameters of the vessel itself. The predicted trajectory of the target vessel is then transformed into a relative motion coordinate system, and the least squares method is used to fit the relative motion trajectory curve. The meeting parameters of the two ships are calculated and output based on the relative motion trajectory curves. Uncertainty analysis is performed on the future navigation trajectory of the target ship and the encounter parameters, and the corresponding confidence interval is output based on the error propagation characteristics of the state estimate. The specific formula for the ship dynamic prediction model is as follows: ； in, The coordinates of the ship in the east direction are: The coordinates of the ship in the north direction. The bow angle of the ship. For the longitudinal speed of the ship, For the ship's lateral speed, The angular velocity of the ship's bow turn. For wind speed, Wind direction angle For correction factor, Indicates the direction of the ocean current.

2. The intelligent collision avoidance decision-making method for ships as described in claim 1, characterized in that: Based on the encounter parameters, a risk assessment matrix is ​​established, and a fuzzy neural network algorithm is used to determine the risk level of the current navigation situation, including: Construct a risk assessment matrix and classify the risk assessment matrix into levels based on the threshold range of the meeting parameters, wherein the meeting parameters include meeting time and meeting distance; Based on the risk assessment matrix, a risk assessment model is constructed using a fuzzy neural network, wherein the fuzzy neural network includes an input layer, a fuzzy layer, a rule layer, and an output layer. The risk assessment model is trained based on historical collision avoidance case data, and the model weight parameters are optimized and the membership function is adjusted through the backpropagation algorithm. Input the current navigation situation data into the trained risk assessment model, calculate the membership value of the collision risk, and output the risk level of the current situation according to the principle of maximum membership. Based on the current risk level, navigation situation data will be continuously collected, and the risk level assessment results will be adjusted and updated in real time.

3. The intelligent collision avoidance decision-making method for ships as described in claim 2, characterized in that: The risk assessment matrix is ​​divided as follows: When the meeting time is less than the first time threshold and the meeting distance is less than the first distance threshold, the corresponding emergency danger level in the risk assessment matrix is: When the meeting time is less than the second time threshold and the meeting distance is less than the second distance threshold, the corresponding risk assessment matrix indicates a high risk level. When the meeting time is between the second and third time thresholds and the meeting distance is between the second and third distance thresholds, it corresponds to a medium risk level in the risk assessment matrix. When the encounter time is greater than the third time threshold or the encounter distance is greater than the third distance threshold, the corresponding risk assessment matrix is ​​classified as low risk level. When the meeting time is greater than the fourth time threshold and the meeting distance is greater than the fourth distance threshold, the corresponding safety level of the risk assessment matrix is ​​determined.

4. The intelligent collision avoidance decision-making method for ships as described in claim 3, characterized in that: The radar data includes the echo signal and distance information of the target vessel; the AIS information includes the latitude and longitude positions, speed values, and heading angles of the vessel and the target vessel.

5. A ship intelligent collision avoidance decision-making system, based on the ship intelligent collision avoidance decision-making method according to any one of claims 1 to 4, characterized in that: include, The data acquisition module is used to collect radar data and AIS information of this vessel; The target ship trajectory prediction module is used to construct a ship dynamic prediction model based on the radar data and AIS information, predict the future navigation trajectory of the target ship through the Kalman filter algorithm, and calculate the encounter parameters of the two ships in combination with the navigation parameters of the ship itself. The judgment module is used to establish a risk assessment matrix based on the encounter parameters and use a fuzzy neural network algorithm to judge the risk level of the current navigation situation. The navigation risk assessment module calculates collision avoidance maneuver parameters based on the risk level assessment results and outputs a collision avoidance action plan using an adaptive particle swarm optimization algorithm.

6. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, it implements the steps of the intelligent collision avoidance decision-making method for ships according to any one of claims 1 to 4.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, it implements the steps of the intelligent collision avoidance decision-making method for ships as described in any one of claims 1 to 4.

Images