A method and system for quantifying collision risk in multiple encounter scenarios for inland waterway vessels

Abstract

The application discloses a kind of inland river ship multi-encounter situation collision risk quantification method and system, obtain the navigation state parameter of ship and target ship, calculate five indexes of minimum encounter distance, minimum encounter time, current distance between ships, relative azimuth and ship speed ratio;With the length of ship as the benchmark, the distance index is dimensionless, and the size difference of the ship is eliminated.According to the relative azimuth and the relationship of heading, the current encounter situation is identified, and the corresponding model is adaptively selected from three sub-models of crossing encounter, overtaking and meeting for reasoning.The three sub-models have different membership function threshold parameters and fuzzy rule bases, and realize differentiated evaluation according to the situation.At the same time, the minimum encounter time allows to take negative value to represent that the two ships have passed the nearest encounter point, and sets residual risk judgment rules in the rule base, which overcomes the sudden drop of risk after encounter caused by the traditional model that takes the minimum encounter time less than zero as the basis for risk removal.

Description

Technical Field

[0001] This invention relates to the field of ship navigation safety technology, and in particular to a method and system for quantifying collision risks in multi-encounter situations for inland waterway vessels. Background Technology

[0002] Inland waterway transportation is an important component of the comprehensive transportation system. With the continuous expansion of inland waterway scale, the navigation density of waterways has significantly increased, leading to increasingly complex encounter relationships between vessels. Inland waterways are characterized by limited space, dense traffic flow, numerous winding sections, and numerous confluences of main and tributary waterways, significantly increasing the collision risk faced by vessels during navigation. In the field of vessel collision risk assessment, Distance to Closest Point of Approach (DCPA) and Time to Closest Point of Approach (TCPA) are the two most widely used geometric encounter parameters. Early studies and maritime research largely based collision risk judgment criteria on these two parameters, assessing the level of collision risk by calculating the minimum encounter distance between two vessels and the time to reach the minimum encounter point. Building on this, subsequent research has gradually introduced methods such as vessel domain models and fuzzy comprehensive evaluation, using factors such as DCPA, TCPA, inter-vessel distance, relative bearing, and speed ratio as evaluation parameters. Fuzzy quantification is used to determine the collision risk level between vessels.

[0003] However, existing technologies still have the following shortcomings. First, the geometric parameter method based on DCPA and TCPA mainly relies on the geometric proximity relationship between the two ships to judge risk, and has limited ability to respond to differences in ship size, avoidance responsibility relationships, and environmental disturbances. Ships of different sizes have significant risk differences under the same DCPA value, but existing methods are unable to reflect this. Second, although the ship domain model can enhance the expression of spatial occupancy relationships, it still has shortcomings in terms of the adaptability of domain parameters under different inland waterway conditions, the reasonable setting of dynamic boundaries, and the coupling of environmental disturbances. Third, although the fuzzy logic method is suitable for describing the fuzziness of the driver's experience judgment, existing models usually use a uniform index system to evaluate different encounter situations indiscriminately, lacking differentiated modeling for scenarios such as cross encounters, overtaking, and head-on encounters, resulting in the need to improve the adaptability and accuracy of the model in the variable inland waterway environment. Summary of the Invention

[0004] Based on the above problems, embodiments of the present invention provide a method and system for quantifying collision risks in multiple encounter situations of inland waterway vessels, in order to solve the technical problems in the prior art that lack the ability to quantitatively assess collision risks in different scenarios for different encounter situations in inland waterways, and cannot characterize the evolution process of residual risks throughout the entire encounter cycle of vessels.

[0005] In a first aspect, embodiments of the present invention provide a method for quantifying collision risk in multi-encounter situations involving inland waterway vessels, the method comprising: Obtain the navigation status parameters of the ship and the target ship, including the ship's position, speed, heading and length; Based on the navigation status parameters, calculate the collision risk assessment index between the ship and the target ship. The collision risk assessment index includes the minimum encounter distance, the minimum encounter time, the current ship-to-ship distance, the relative bearing angle, and the ship speed ratio. Using the ship's length as a dimensionless reference scale, the minimum meeting distance and the current ship-to-ship distance are processed to obtain the dimensionless minimum meeting distance and the dimensionless ship-to-ship distance. Based on the current encounter situation between the vessel and the target vessel, a target sub-model corresponding to the current encounter situation is determined from multiple pre-constructed encounter situation sub-models. These multiple encounter situation sub-models include a cross-encounter sub-model, an overtaking sub-model, and a face-to-face sub-model. Each encounter situation sub-model uses a Mamdani-type IF-THEN rule to construct an independent fuzzy rule base, and each of the multiple encounter situation sub-models has a unique membership function threshold parameter configuration. Each independent fuzzy rule base is also unique. When the minimum encounter time is negative, it indicates that the two vessels have passed the nearest encounter point. The dimensionless minimum encounter distance, the minimum encounter time, the dimensionless inter-ship distance, the relative azimuth angle, and the ship speed ratio are input into the target sub-model, and the target sub-model outputs the collision risk level based on fuzzy logic reasoning.

[0006] Secondly, embodiments of the present invention provide a collision risk quantification system for inland waterway vessels in multi-encounter situations, the system comprising: The navigation status parameter acquisition module is used to acquire the navigation status parameters of the ship and the target ship, including the ship's position, speed, heading and length. The evaluation index calculation module is connected to the navigation state parameter acquisition module and is used to calculate the collision risk evaluation index between the ship and the target ship based on the navigation state parameters. The collision risk evaluation index includes minimum encounter distance, minimum encounter time, current ship-to-ship distance, relative bearing angle and ship speed ratio. The dimensionless processing module, connected to the evaluation index calculation module, is used to perform dimensionless processing on the minimum encounter distance and the current inter-ship distance using the ship's length as the dimensionless benchmark scale, to obtain the dimensionless minimum encounter distance and the dimensionless inter-ship distance. The meeting situation sub-model storage module is used to store multiple pre-built meeting situation sub-models, including cross-meeting sub-models, overtaking sub-models, and face-to-face meeting sub-models; the multiple meeting situation sub-models have different membership function threshold parameter configurations and different fuzzy rule bases; when the minimum meeting time is negative, it indicates that the two ships have passed the nearest meeting point; The encounter situation identification and sub-model selection module is connected to the dimensionless processing module and the encounter situation sub-model storage module. It is used to determine the target sub-model corresponding to the current encounter situation from the encounter situation sub-model storage module based on the current encounter situation between the ship and the target ship. The fuzzy reasoning module, connected to the dimensionless processing module and the encounter situation recognition and sub-model selection module, is used to input the dimensionless minimum encounter distance, the minimum encounter time, the dimensionless inter-ship distance, the relative azimuth angle and the ship speed ratio into the target sub-model, and output the collision risk level based on fuzzy logic reasoning.

[0007] This invention discloses a method and system for quantifying collision risk in multi-encounter situations for inland waterway vessels. First, it acquires navigational parameters such as the vessel's position, speed, course, and length, as well as those of the target vessel, and calculates five collision risk assessment indicators based on these parameters: minimum encounter distance, minimum encounter time, current distance between vessels, relative bearing, and speed ratio. Second, it uses the vessel's length as a dimensionless reference scale to perform dimensionless processing on the minimum encounter distance and current distance between vessels, thereby eliminating the influence of vessel size differences on risk assessment. Based on this, it identifies the current encounter based on the relative bearing and course relationship between the vessel and the target vessel. The invention analyzes the situation and adaptively selects the corresponding target sub-model from three pre-constructed sub-models: a cross-encounter sub-model, a overtaking sub-model, and a face-to-face sub-model. Each of these three sub-models has distinct membership function threshold parameter configurations and fuzzy rule bases, enabling differentiated assessment of different encounter scenarios. In particular, the invention allows the minimum encounter time to be negative to indicate that both vessels have passed the nearest encounter point, and sets corresponding residual risk judgment rules in the fuzzy rule base. This overcomes the defect of traditional models that use a near-zero nearest encounter time as the basis for risk resolution, leading to a sudden drop in risk after the encounter. Through these technical means, the invention can continuously and smoothly quantify the collision risk of inland waterway vessels throughout the entire encounter process. It solves the problems of existing technologies, such as the lack of a unified comparison benchmark for vessels of different scales and the lack of scenario differentiation capabilities in the geometric parameter method, and overcomes the technical deficiency of traditional models in failing to characterize residual risk after an encounter. This provides a more accurate and reliable risk warning basis for the navigation safety of inland waterway vessels. Attached Figure Description

[0008] Figure 1A flowchart of a method for quantifying collision risk in multi-encounter situations for inland waterway vessels provided in an embodiment of the present invention; Figure 2 This is a schematic diagram of the scenario-specific differentiated modeling architecture provided in an embodiment of the present invention; Figure 3 A schematic diagram of the membership functions of five key indicators of the CR sub-model provided in this embodiment of the invention; Figure 4 This is a risk comparison diagram between the sub-models provided in the embodiments of the present invention and other models in the prior art; Figure 5 The structural block diagram of the collision risk quantification system for inland waterway vessels in multi-encounter situations provided in this embodiment of the invention. Detailed Implementation

[0009] Figure 1 This is a flowchart illustrating a method for quantifying collision risk in multi-encounter situations for inland waterway vessels according to an embodiment of the present invention. (Refer to...) Figure 1 The method includes: S1. Obtain the navigation status parameters of the ship and the target ship, including the ship's position, speed, heading and length.

[0010] In this embodiment, S1, which obtains the navigation status parameters of the ship and the target ship, is the data basis for collision risk quantification. The navigation status parameters include the ship's position, speed, heading, and length.

[0011] Among them, the ship's position is used to determine the spatial relative relationship between the two ships, the speed and heading are used to describe the ship's motion trend, and the ship's length serves as the reference scale for subsequent dimensionless processing.

[0012] In actual inland waterway navigation scenarios, the above parameters can be obtained in real time through equipment such as Automatic Identification System (AIS) or Automatic Radar Plotting Aid (ARPA).

[0013] S2. Based on the navigation status parameters, calculate the collision risk assessment index between the ship and the target ship. The collision risk assessment index includes the minimum encounter distance, the minimum encounter time, the current ship-to-ship distance, the relative bearing angle, and the ship speed ratio.

[0014] In this embodiment, S2 calculates the collision risk assessment index between the ship and the target ship based on the acquired navigation state parameters. The assessment index includes five core indicators: minimum encounter distance (Distance to Closest Point of Approach, DCPA), minimum encounter time (TCPA), current ship-to-ship distance D, relative bearing angle β, and ship speed ratio K.

[0015] Among them, the minimum encounter distance (DCPA) is used to characterize the safety margin when the two ships reach the nearest encounter point; the smaller the value, the higher the collision risk. The minimum encounter time (TCPA) is used to reflect the urgency of the time when the two ships reach the nearest encounter point; the shorter the value, the less time the driver has to make judgments and maneuvers. The current ship-to-ship distance (D) is used to describe the current absolute distance between the two ships, reflecting whether there is a risk of distance pressure or ship suction effect. The relative bearing angle (β) is used to describe the bearing of the target ship relative to the ship, providing a basis for determining the encounter angle and avoidance responsibility. The speed ratio (K) is used to characterize the speed difference between the two ships, reflecting speed pressure and approach trend.

[0016] S3. Using the length of this ship as a dimensionless reference scale, perform dimensionless processing on the minimum meeting distance and the current ship-to-ship distance to obtain the dimensionless minimum meeting distance and the dimensionless ship-to-ship distance.

[0017] It should be noted that in inland waterway navigation practice, there are significant differences in the types and sizes of vessels. Large vessels and small vessels face completely different actual risks at the same absolute distance. If absolute distance units such as meters and nautical miles are used directly for risk assessment, there is a lack of a unified comparison benchmark between vessels of different sizes, which may lead to a large deviation between the assessment results and the actual situation.

[0018] In this embodiment, S3 transforms the spatial dimension index from absolute distance to a relative value in units of the ship's length through dimensionless processing, providing a unified comparison benchmark for ships of different sizes in risk assessment. This step uses the ship's length as the dimensionless benchmark scale to perform dimensionless processing on the minimum encounter distance DCPA and the current inter-ship distance D, obtaining dimensionless minimum encounter distance DCPA* and dimensionless inter-ship distance D*, respectively. This eliminates the influence of differences in ship physical dimensions on risk judgment, making the model universal and applicable to encounter situations involving different ship types in inland waterways.

[0019] S4. Based on the current encounter situation between the ship and the target ship, determine the target sub-model corresponding to the current encounter situation from multiple pre-constructed encounter situation sub-models. The multiple encounter situation sub-models include cross-encounter sub-models, overtaking sub-models, and face-to-face sub-models. The multiple encounter situation sub-models have different membership function threshold parameter configurations and different fuzzy rule bases. When the minimum encounter time is negative, it indicates that the two ships have passed the nearest encounter point.

[0020] Specifically, in this embodiment, S4 identifies the current encounter situation as a cross encounter, overtaking, or head-on encounter based on the relative bearing and heading relationship between the vessel and the target vessel, and selects one of the pre-built cross encounter sub-models, overtaking sub-models, and head-on encounter sub-models as the target sub-model. Since the collision risk evolution under different encounter situations such as cross encounters, overtaking, and head-on encounters of inland waterway vessels has its own characteristics, the danger of cross encounters mainly comes from track intersections and differences in avoidance responsibilities, the danger of overtaking mainly comes from the continuous approach and close-range paralleling of the following vessel, and the danger of head-on encounters mainly comes from high relative speeds and short meeting times. Therefore, it is necessary to construct a dedicated sub-model for each encounter situation.

[0021] The multiple encounter situation sub-models have different membership function threshold parameter configurations and different fuzzy rule bases: the membership function threshold parameter configuration determines the membership degree of the same input variable value in different sub-models, and the fuzzy rule base determines how the combination of multiple indicators is mapped to the risk level; both are different among the three sub-models, thus ensuring that the same encounter parameter can produce differentiated risk judgment results under different encounter situations.

[0022] In particular, the minimum meeting time TCPA is allowed to be negative. When TCPA is less than zero, it means that the two ships have passed the nearest meeting point. At this time, the model does not simply reduce the risk to zero, but still needs to combine indicators such as dimensionless inter-ship distance D* and dimensionless minimum meeting distance DCPA* to determine whether the two ships are still in close range, so as to retain the ability to judge residual risks such as ship suction effect and wake interference.

[0023] This step achieves accurate scenario-based assessment of different encounter situations for inland waterway vessels and a complete characterization of the risk throughout the entire encounter cycle by pre-constructing multiple encounter sub-models with different membership function threshold parameter configurations and different fuzzy rule bases, and by introducing a residual risk judgment mechanism corresponding to the negative value of TCPA.

[0024] S5. Input the dimensionless minimum encounter distance, the minimum encounter time, the dimensionless inter-ship distance, the relative azimuth angle, and the ship speed ratio into the target sub-model. The target sub-model outputs the collision risk level based on fuzzy logic reasoning.

[0025] Specifically, the five input variables—dimensionalized minimum encounter distance DCPA*, minimum encounter time TCPA, dimensionless inter-ship distance D*, relative bearing angle β, and ship speed ratio K—are input into the target sub-model determined in step S4. The target sub-model outputs the Collision Risk Index (CRI) based on fuzzy logic reasoning.

[0026] The fuzzy logic reasoning process is as follows: First, each input variable is transformed into a membership value corresponding to each fuzzy linguistic variable through a trapezoidal membership function, converting precise numerical inputs into fuzzy linguistic descriptions. Second, based on the constructed Mamdani-type IF-THEN fuzzy rule base, the fuzzy linguistic state combination of each input variable is correlated with the output risk level. The rule base design incorporates the operational experience of senior ship operators and the requirements of the "Rules for Avoiding Collisions on Inland Waterways of the People's Republic of China" regarding ship responsibility, avoidance timing, and degree of danger under different encounter situations. Finally, through fuzzy reasoning synthesis and centroid method defuzzification calculation, a continuous collision risk index (CRI) value between 0 and 1 is output. This value represents the collision risk, and the higher the value, the greater the collision risk.

[0027] Based on the above embodiments, as a preferred implementation, multiple encounter situation sub-models all use the dimensionless minimum encounter distance, the minimum encounter time, the dimensionless inter-ship distance, the relative azimuth angle, and the ship speed ratio as input variables, and the collision risk as the output variable.

[0028] Each of the aforementioned encounter scenario sub-models employs a trapezoidal membership function to characterize the membership degree of each input variable under different risk levels. Each input variable is divided into multiple fuzzy linguistic variable intervals, with overlapping regions between adjacent intervals. In the cross-meeting sub-model, the fuzzy linguistic variable interval for the minimum meeting time includes the interval of the first nearest meeting point that has passed, the interval of the first time urgency state, the interval of the first medium urgency state, and the interval of the first time sufficiency state; the fuzzy linguistic variable interval for the relative azimuth angle includes the interval of the first right direction, the interval of the first stern direction, and the interval of the first left direction.

[0029] In the overtaking sub-model, the fuzzy linguistic variable interval for the minimum encounter time includes the interval of the second nearest encounter point that has been passed, the interval of the second time urgency state, the interval of the second medium urgency state, and the interval of the second time sufficiency state; the fuzzy linguistic variable interval for the relative azimuth angle includes the interval of the second right front, the interval of the second right horizontal direction, the interval of the second right rear, the interval of the second left rear, the interval of the second left horizontal direction, and the interval of the second left front.

[0030] In the encounter sub-model, the fuzzy linguistic variable interval for the minimum encounter time includes the interval for the third most recent encounter point that has passed, the interval for the third time-urgent state, the interval for the third medium-urgent state, and the interval for the third time-sufficient state; the fuzzy linguistic variable interval for the relative azimuth angle includes the interval for the third right front, the interval for the third right side, the interval for the third left side, and the interval for the third left front.

[0031] In a preferred embodiment of the present invention, each of the multiple encounter situation sub-models uses dimensionless minimum encounter distance DCPA*, minimum encounter time TCPA, dimensionless inter-ship distance D*, relative bearing angle β, and ship speed ratio K as input variables, and collision risk index (CRI) as output variable. Each encounter situation sub-model employs a trapezoidal membership function to characterize the membership degree of each input variable at different risk levels.

[0032] Each input variable is divided into multiple fuzzy linguistic variable intervals, with overlapping areas between adjacent intervals. Using a segmented, overlapping trapezoidal membership function ensures that the model does not abruptly change when a parameter just crosses a threshold, but rather transitions smoothly between adjacent states. The mathematical expression for the trapezoidal membership function is:

[0033] In the formula, x is the value of the input variable, and a, b, c, and d are the fuzzy threshold boundaries of each input variable under different collision risk levels.

[0034] The output variable CRI of the three sub-models all adopts a Gaussian membership function. The universe of discourse of CRI is [0,1]. Four fuzzy linguistic variables, namely Low risk, Mid risk, High risk, and VHigh risk, are selected as the evaluation set of ship navigation risk.

[0035] In this invention, the universe of discourse for each input variable is defined by a trapezoidal membership function. The universe of discourse for each input variable in each sub-model is shown in the membership threshold parameter table (e.g., the universe of discourse for TCPA in Table 1 is [-30, 60]). A fuzzy linguistic variable interval refers to dividing the continuous universe of discourse of an input variable into several qualitatively descriptive intervals with actual physical meaning. Each input variable is divided into multiple fuzzy linguistic variable intervals. For example, TCPA is divided into "Past" interval, "Time-pressured interval" (S), "Medium-pressured interval" (M), and "Time-sufficient interval" (L). The fuzzy linguistic variable intervals for each input variable and their corresponding threshold parameters are all preset. Here, S, M, and L represent Small, Middle, and Large, respectively. The small value range indicates that the input variable's value is within a small range, corresponding to a state of "small space margin" or "tight time"; the medium value range indicates a moderate value; and the large value range indicates a large value, corresponding to a state of "large space margin" or "ample time". The "Past" interval is a fuzzy linguistic variable unique to TCPA, indicating that TCPA is less than 0, meaning that the two ships have passed the nearest meeting point. The universe of discourse for the output variable Collision Risk Ratio (CRI) of each sub-model is [0,1], defined by a Gaussian membership function, and four fuzzy linguistic variables—Low (low risk), Mid (medium risk), High (high risk), and VHigh (extremely high risk)—are selected as risk levels.

[0036] "The interval ××× is valid" means that the precise value of an input variable, after being mapped by the trapezoidal membership function, has a membership degree greater than 0 for the interval of the fuzzy linguistic variable; conversely, a membership degree of 0 indicates that the interval is invalid. When multiple intervals are valid simultaneously, the corresponding fuzzy rule is activated. For example, "the interval of the smallest dimensionless inter-ship distance is valid" means that the precise value of D* has a membership degree greater than 0 for the interval S, i.e., D* is determined to belong to the smallest interval.

[0037] (I) Construction of Crossing (CR) Sub-model In the cross-encounter sub-model, the fuzzy linguistic variable intervals for the minimum encounter time TCPA include the first passed nearest encounter point interval Past, the first time urgency interval S, the first moderate urgency interval M, and the first time sufficiency interval L. The fuzzy linguistic variable intervals for the relative azimuth angle β include the first right interval Right, the first stern direction interval Stern, and the first left interval Left.

[0038] In the cross-encounter sub-model, TCPA introduces a Past state to describe the risk decay process after the two vessels have passed the nearest meeting point, avoiding the traditional model's direct reduction of risk to zero after TCPA < 0. For cross-encounters in inland waterways where tributaries meet or in meandering waterways, the nearest meeting point cannot be simply regarded as the dividing line where risk ends. Although the two vessels have passed the CPA, if D* is still small at this point, it indicates that the actual distance between the two vessels has not yet fully increased, and wake interference, ship suction effects, and the pressure on the pilots to make subsequent maneuvering adjustments may still exist. Therefore, the model still needs to retain a certain risk output.

[0039] The design of the relative bearing angle β takes into account the actual situation of oncoming vessel bearing, avoidance responsibility, close passage, and residual risks after passing in inland waterway cross-encounters. In cross-encounter situations, when the target vessel comes from the starboard side, the vessel usually faces more direct avoidance pressure. Therefore, in the rule base design, if the target vessel is in the Right area and has a small DCPA*, short TCPA, and close D*, the model will give a higher danger level.

[0040] The membership threshold parameters of the five key indicators of the cross-meeting sub-model under different risk levels are shown in Table 1.

[0041] Table 1. Membership threshold parameters of five key indicators in the CR sub-model at different risk levels.

[0042] In the table above, S, M, and L represent Small (small value range), Middle (medium value range), and Large (large value range), respectively, corresponding to small, medium, and large values ​​of the input variable; Past indicates that the ship has passed the nearest meeting point when TCPA is less than 0. For the relative bearing angle β, Right, Stern, and Left represent the target ship being to the right, stern, and left of the ship, respectively. HO1 and HO2 both represent the encounter direction. Since the domain of the relative bearing angle β is 0° to 360°, and the encounter direction is near 0°, this embodiment divides it into two intervals, HO1 and HO2, to ensure that the membership function can continuously describe the angle boundary.

[0043] like Figure 3 As shown, Figure 3-1 to Figure 3-5The membership function curves for the five input variables DCPA*, TCPA, K, D*, and β are displayed respectively. This type of diagram can intuitively illustrate two important design features of the present invention: First, there is an overlapping region between adjacent fuzzy linguistic variable intervals, which allows the model to smoothly transition when parameters cross interval boundaries, avoiding abrupt changes in risk values; second, the TCPA membership function includes a Past state interval (corresponding to TCPA<0), enabling the model to retain residual risk judgment capabilities by combining indicators such as D* after the ship has passed the nearest meeting point. This is one of the key technical features that distinguishes the present invention from traditional models.

[0044] The fuzzy logic rule base for the cross-encounter sub-model adopts the Mamdani-type IF-THEN rule format, with a total of 41 rules. The rule design closely reflects the basic thinking of the driver when judging danger: first, it determines whether there is a clear trend of approach between the two ships based on DCPA and TCPA, then it determines whether the danger has developed to the close-range stage based on the current D*, then it determines the location of the target ship through β, and finally uses K to make appropriate corrections to the degree of danger.

[0045] When the target vessel is located on the starboard side of the vessel, and the DCPA* is small, the TCPA is short, and the D* is close, the rule determines it to be of a high or even very high risk level, reflecting the requirement for the vessel to take early evasive action when acting as the give-way vessel. For targets approaching from the port side or stern direction, the judgment is further made in conjunction with the distance, time, and speed relationships; if the two vessels have already formed close proximity, the risk level will be increased accordingly. When both the first interval of the nearest encounter point and the minimum interval of the dimensionless inter-vessel distance are met simultaneously, the rule base retains a medium or high risk output, thereby achieving a continuous characterization of the residual risk after the vessel has passed the nearest encounter point. Among them, "simultaneously in the corresponding fuzzy language state" means that each input variable is simultaneously determined to be the value of the corresponding fuzzy language variable.

[0046] The fuzzy logic rule base of the cross-meeting sub-model adopts the Mamdani type IF-THEN rule form, and sets up multiple fuzzy rules, as shown in Table 2.

[0047]

[0048] Note: In the table, "-" indicates any state; multiple values ​​separated by " / " in the same cell indicate that the row contains multiple parallel rules, and the corresponding variable of each rule takes the corresponding value separated by " / "; an empty cell indicates that there is no rule with a value of "-" in the corresponding column of the row. For example, "S / S / S" means that the variable takes the values ​​S, S, and S in the three rules respectively. "H" indicates that the ship speed ratio K is in the large range. "Past" indicates that the minimum meeting time TCPA is in the range where the nearest meeting point has passed, i.e., TCPA < 0.

[0049] The rule base takes five variables as input: dimensionless minimum encounter distance (DCPA*), minimum encounter time (TCPA), dimensionless inter-ship distance (D*), relative bearing angle (β), and ship speed ratio (K), and outputs the collision risk level (CRI). The rule design first determines the spatial safety margin when the two ships reach the nearest encounter point based on the fuzzy linguistic state of the dimensionless minimum encounter distance (DCPA*), and simultaneously determines the time urgency of the two ships reaching the nearest encounter point based on the fuzzy linguistic state of the minimum encounter time (TCPA). If both DCPA* and TCPA are in a low value range, it indicates a clear trend of approaching between the two ships. Based on this, the fuzzy linguistic state of the dimensionless inter-ship distance (D*) is used to determine whether the current distance between the two ships has entered the close proximity stage. Then, the fuzzy linguistic state of the relative bearing angle (β) is used to determine the location of the target ship. Finally, the fuzzy linguistic state of the ship speed ratio (K) is used to correct the risk level.

[0050] The cross-meeting sub-model sets up multiple fuzzy rules, and the core judgment logic revolves around two key factors: "location responsibility" and "spatial urgency".

[0051] The logical thread of the rule design is as follows: First, determine whether there is a clear trend of approach between the two ships based on DCPA* and TCPA. Then, combine the current D* to determine whether the danger has developed to the close range stage. Next, determine the location of the target ship through β. Finally, use K to correct the danger level.

[0052] The key judgment rules can be summarized as follows: Extremely High Risk Level Rule: When β is in the first right-hand zone, and DCPA*, TCPA, and D* are all in the S zone, the risk level is determined to be extremely high. This reflects the requirement that the vessel, acting as the give-way vessel, must take evasive action as early as possible.

[0053] High-risk level rule: When β is in the first right interval and both DCPA* and D* are in the S / M interval, or when β is in the first left interval and both DCPA* and TCPA are in the S interval, it is determined to be a high-risk level.

[0054] Past State Rule: When TCPA is in the Past interval and D* is still in the S interval, even if the two ships have passed the nearest meeting point, the rule base still retains the output of medium or high risk level, reflecting the characterization of residual risk.

[0055] Specifically, when the relative azimuth angle β is in the first right interval, and the dimensionless minimum encounter distance DCPA* is in a small interval (S), the minimum encounter time TCPA is in a small interval (S), and the dimensionless inter-ship distance D* is in a small interval (S), the rule base determines the collision risk level to be extremely high, so as to reflect the requirement that the vessel should take early avoidance action when acting as the give-way vessel.

[0056] When the relative azimuth angle β is in the first left interval or the first stern interval, the rule base continues to make a comprehensive judgment based on the fuzzy language state of the dimensionless minimum encounter distance DCPA*, minimum encounter time TCPA, dimensionless inter-ship distance D*, and ship speed ratio K; if the two ships have formed a preset close-range pressure relationship, the danger level is increased accordingly.

[0057] When the fuzzy linguistic variable of TCPA is in the first "Past" interval, it indicates that the two ships have passed the nearest meeting point. However, if the dimensionless inter-ship distance D* is still in the small value interval (S) at this time, the rule base retains the output of medium or high risk level to reflect the residual risks that may still exist after the two ships pass close together, such as ship suction effect, wake interference, and pilot control adjustment pressure. As the inter-ship distance gradually increases to the medium or large value interval, the risk level decreases accordingly.

[0058] (II) Construction of the Overtaking (OT) Model In the overtaking sub-model, the fuzzy linguistic variable intervals for the minimum encounter time TCPA include the second passed nearest encounter point interval Past, the second time-urgent state interval S, the second moderately urgent state interval M, and the second time-sufficient state interval L. The fuzzy linguistic variable intervals for the relative azimuth angle β include the second right front interval Front_R, the second right lateral direction region interval Side_R, the second right rear interval Stern_R, the second left rear interval Stern_L, the second left lateral direction region interval Side_L, and the second left front interval Front_L.

[0059] In the overtaking sub-model, the membership function is designed primarily around the actual overtaking process in inland waterways. Overtaking differs from head-on encounters and cross encounters; it often lasts longer, involves lower relative speeds, and the two vessels may maintain close following or parallel travel for an extended period. Therefore, the overtaking sub-model employs a trapezoidal membership function with overlapping intervals to avoid sudden jumps in risk level when parameters just cross a boundary. The S interval of DCPA* is set quite sensitively because a small lateral distance during overtaking not only means the physical distance between the two vessels is close but may also lead to problems such as ship suction, wake interference, and limited maneuverability. D* is used to describe the current distance between the two vessels, compensating for the difficulty of relying solely on DCPA* to reflect the collision risk during close-range parallel travel. The rule base shows that the overtaking model is particularly sensitive to the combination of "small DCPA* and close D*," which aligns with the overtaking situation in inland waterways. When vessels are traveling in the same direction, even if a direct collision does not occur in a short time, if the lateral distance is too small and the distance is too close, the driver will still experience significant risk pressure. K is mainly used to describe the speed difference between the two vessels. When an overtaking vessel has a significant speed advantage, the following vessel will continuously reduce the safety margin between itself and the preceding vessel, and the risk should be reflected in advance. In addition to S, M, and L, TCPA also sets a Past state, which is also very important. After an overtaking vessel passes the nearest encounter point, it does not mean that the risk is immediately eliminated. The ship suction effect and wake effect may still persist, therefore the rule base will retain a certain amount of risk output.

[0060] The design of the relative bearing angle β reflects the changing characteristics of an overtaking situation. The model divides β into six regions: starboard forward, starboard beam, starboard aft, starboard aft, starboard beam, and starboard forward, to describe the relative position changes of the target vessel during the overtaking process. Overtaking is not an instantaneous process; the target vessel may gradually approach from the stern of the main vessel, then enter the beam position, and then gradually move towards the bow. The bearing changes themselves can reflect the progression of the overtaking process.

[0061] The membership threshold parameters of the five key indicators of the chasing sub-model at different risk levels are shown in Table 3.

[0062] Table 3. Membership threshold parameters of five key indicators of the OT sub-model at different risk levels.

[0063] In the table above, Front_R, Side_R, Stern_R, Stern_L, Side_L and Front_L respectively indicate that the target vessel is located in the area to the right front, to the right transverse, to the right rear, to the left rear, to the left transverse, and to the left front of the vessel.

[0064] The fuzzy logic rule base of the overtaking sub-model (OT) adopts the Mamdani-type IF-THEN rule form. It takes five variables as input: dimensionless minimum encounter distance (DCPA*), minimum encounter time (TCPA), dimensionless inter-ship distance (D*), relative bearing angle (β), and ship speed ratio (K). The output is the collision risk level (CRI). Multiple fuzzy rules are set, and the specific rules are shown in Table 4. The rule base revolves around several factors that are most likely to cause danger during overtaking: whether the following ship continuously approaches the preceding ship, whether D* is too close, whether DCPA* is too small, and whether the target ship is within a suitable bearing range for overtaking. When the following ship has a significant speed advantage and the distance is relatively close, the danger will gradually increase even if the bearing angle does not change much.

[0065] The rule base introduces the Past state of TCPA to describe the situation where two ships have passed the nearest meeting point. When the fuzzy linguistic variable of TCPA is in the second Past meeting point interval (Past), it indicates that the two ships have passed the nearest meeting point. However, if the dimensionless inter-ship distance D* is still in the small value interval (S) or the dimensionless minimum meeting distance DCPA* is still in the small value interval (S), the rule base retains the output of medium or high risk level to reflect the residual risks that may still exist after the two ships pass close together or run parallel, such as ship suction effect, wake interference, and limited maneuverability. When both D* and DCPA* have widened to the medium or large value interval, the risk level is reduced accordingly.

[0066] Table 4. Fuzzy logic rule base for different risk levels of the OT sub-model

[0067] In the table, "-" indicates any state (Any); multiple values ​​separated by " / " in the same cell indicate that the row contains multiple parallel rules, and the corresponding variable of each rule takes the corresponding value separated by " / "; an empty cell indicates that there is no rule with the value "-" in the corresponding column of the row. For example, "S / S / S" means that the variable takes the values ​​S, S, and S in the three rules respectively. "SR" represents the second right rear interval (Stern_R), "SL" represents the second left rear interval (Stern_L), "Side_R" represents the second right horizontal region interval, "Side_L" represents the second left horizontal region interval, and "Front_R" represents the second right front interval. "Past" indicates that the minimum encounter time TCPA is in the interval after the nearest encounter point has passed, i.e., TCPA < 0.

[0068] The rule base revolves around several factors that are most likely to cause danger during overtaking: whether the following vessel continues to approach the preceding vessel, whether the D* (difference distance) is too close, whether the DCPA* (direction of approach distance) is too small, and whether the target vessel is within a suitable bearing range for overtaking. When the following vessel has a significant speed advantage and the distance is close, the danger will gradually increase even if the bearing angle does not change much. The rule base introduces the Past state of TCPA (Temporal Capability Aspect) to describe the situation where the two vessels have passed the nearest encounter point, preventing the model from continuing to classify a target that has already begun to drift away as high-risk. However, the Past state does not mean that the danger has completely disappeared; it still needs to be judged in conjunction with the current D* and DCPA*: if the two vessels are still within close range, especially if the DCPA* is small, there may still be a risk of ship suction or limited maneuverability; if the distance has increased, the danger level can be significantly reduced.

[0069] The overtaking sub-model sets up multiple fuzzy rules, and the core judgment logic revolves around two key factors: "continuous approach" and "speed pressure".

[0070] The main logic of the rule design is: whether the following vessel is continuously approaching the preceding vessel, whether the D* is too close, whether the DCPA* is too small, and whether the target vessel is within a suitable bearing range for overtaking.

[0071] The key judgment rules can be summarized as follows: Extremely High Risk Level Rule: When DCPA* is in the S section, D* is in the S section, and β is in the right lateral direction area (Side_R) or the left lateral direction area (Side_L), it is judged as extremely high risk level. This reflects the characteristic that the risk is most concentrated when closely maneuvering during overtaking.

[0072] High-risk level rule: When DCPA* is in the S or M interval, D* is in the S interval, and K has a significant speed advantage, it is judged as a high-risk level.

[0073] Past State Rule: When TCPA is in the Past interval, it is still necessary to combine D* and DCPA* for judgment. If the two ships are still in close range, especially if DCPA* is small, there may still be a risk of ship suction or restricted maneuverability. The rule base retains medium or high risk level output.

[0074] (III) Construction of the Head-on (HO) Model In the aforementioned encounter sub-model, the fuzzy linguistic variable intervals for the minimum encounter time TCPA include the third past nearest encounter point interval Past, the third time-urgent state interval S, the third medium-urgent state interval M, and the third time-sufficient state interval L. The fuzzy linguistic variable intervals for the relative azimuth angle β include the third right front interval Front_R, the third right side interval Side_R, the third left side interval Side_L, and the third left front interval Front_L.

[0075] In the encounter situation sub-model, the membership function is set primarily around the characteristics of inland waterway encounters: "high relative speed and limited meeting space." In an encounter situation, the two vessels typically approach in opposite directions with high relative speeds, leaving the driver with limited time for judgment and maneuvering. Therefore, the S, M, and L interval settings in TCPA emphasize the urgency brought about by time compression. D* is used to describe the current distance between vessels; when the distance has entered the short-range interval, even if other indicators are not entirely extreme, the rule base will increase the risk level.

[0076] The design of the relative azimuth angle β reflects the actual characteristics of the encounter situation. The encounter sub-model divides it into the starboard front, starboard, port, and left front of the ship, focusing on determining whether the target ship is near the bow and whether a head-on approach is forming. For encounters on curves, the target ship may gradually appear from the left or right of the bow, so each boundary interval needs to retain a certain degree of overlap to avoid sudden changes in risk assessment caused by small changes in azimuth angle.

[0077] Table 5 shows the membership threshold parameters of the five key indicators of the encounter model at different risk levels.

[0078]

[0079] In the table above, Front_R, Side_R, Side_L, and Front_L represent the target vessel being located to the right, starboard, port, and left front of the main vessel, respectively. In a head-on encounter situation, when the relative bearing angle β is within the Front_R or Front_L range, it indicates that the target vessel is located near the main vessel's bow direction, and the two vessels are in a head-on or near-head-on encounter situation.

[0080] The fuzzy logic rule base for the encounter sub-model (HO) adopts the Mamdani-type IF-THEN rule form. It takes five variables as input: dimensionless minimum encounter distance DCPA*, minimum encounter time TCPA, dimensionless inter-ship distance D*, relative bearing angle β, and ship speed ratio K, and takes collision risk level CRI as output. Multiple fuzzy rules are set, and the specific rule content is shown in Table 6.

[0081]

[0082] Note: In the table, "-" indicates any state; multiple values ​​separated by " / " in the same cell indicate that the row contains multiple parallel rules, and the corresponding variable of each rule takes the corresponding value separated by " / "; an empty cell indicates that there is no rule with the value "-" in the corresponding column of the row. For example, "S / S / S" means that the variable takes the values ​​S, S, and S in the three rules respectively. "Fast" indicates the qualitative state of "faster speed" when the ship speed ratio K is larger, and its threshold parameter corresponds to the large value range (L) of K in Table 5. "Front_R" indicates the third right front interval, "Front_L" indicates the third left front interval, "Side_R" indicates the third right side interval, and "Side_L" indicates the third left side interval. "Past" indicates that the minimum encounter time TCPA is in the interval after the nearest encounter point has passed, that is, TCPA < 0.

[0083] The rule base is not designed to exhaustively list all possible combinations of variables, but rather to focus on the most critical clues for judgment during an encounter: whether the two ships are in the same heading, whether they are continuously approaching, whether the minimum encounter distance is sufficiently small, and whether the current distance has entered the critical range. Therefore, the rule base makes extensive use of "Any" conditions, highlighting the dominant role of TCPA and D* in determining dangerous changes.

[0084] Specifically, when the minimum encounter time TCPA is in a small range (S) and the dimensionless inter-ship distance D* is in a small range (S), regardless of the state of other indicators, the rule base determines the collision risk level to be extremely high, so as to reflect the extreme danger after the superposition of time and space urgency in the encounter situation.

[0085] When the relative azimuth angle β is in the third right forward interval (Front_R) or the third left forward interval (Front_L) and the dimensionless minimum encounter distance DCPA* is in the small value interval (S), it indicates that the two ships are approaching in the bow direction and the space safety margin is small. The rule base determines it to be a high risk level or an extremely high risk level to reflect the urgency of the head-on approach in the encounter situation.

[0086] When the fuzzy linguistic variable of TCPA is in the third "Past" interval, it indicates that the two ships have passed the nearest meeting point. However, if the dimensionless inter-ship distance D* is still in the small range (S), the rule base determines that the two ships still face the threat of wake interference and ship suction effects, and forces the output to maintain a medium risk level instead of directly resetting it to zero. This design reflects the fact that residual risks still objectively exist after a close encounter in a meeting situation. Although the two ships have passed the nearest meeting point, if the actual distance has not yet fully increased, the hydrodynamic interaction between the ships and the pressure on the pilots to adjust their maneuvers may still continue to affect navigation safety. As the inter-ship distance gradually increases to the medium range (M) or the large range (L), the risk level smoothly decreases to low risk.

[0087] It should be noted that the membership functions of both the OT and HO sub-models employ trapezoidal membership functions, dividing each input variable into fuzzy linguistic variables with the same structure, and also including overlapping regions between adjacent fuzzy linguistic variable intervals. Furthermore, the TCPA membership functions of both the OT and HO sub-models also include the "Past" interval, used to describe the residual risk after the ship has passed the nearest encounter point. The only difference in the membership function design among the three sub-models lies in the domain range of the same input variable and the specific values ​​of the threshold parameters; their core design structure and the achieved technical effects are completely identical.

[0088] like Figure 4 The figure shows the time evolution curves of the collision risk CRI of the embodiment model of the present invention (corresponding to the CR sub-model, OT sub-model and HO sub-model respectively) and two control models, the traditional DCPA / TCPA geometric parameter model (hereinafter referred to as the "traditional model") and the collision risk model based on the four-dimensional ship domain (hereinafter referred to as the "domain model"), under three typical inland waterway encounter situations. In contrast, the CRI curves of the model of this invention exhibit three common characteristics in all three scenarios: First, during the approach phase of the ships, the risk escalation is smoother and without abrupt changes. In the cross-encounter scenario, the CRI begins to rise significantly after about 222 seconds; in the overtaking scenario, it rises rapidly after about 100 seconds and maintains a high risk of around 0.9; and in the encounter scenario, it rises significantly after about 250 seconds and reaches a high risk between 324 and 423 seconds. The escalation pattern in each scenario is continuous and stable. Second, in the high-risk identification phase near the CPA, the model of this invention can provide a reasonable risk value that matches the actual navigation pressure on inland waterways, avoiding the over-warning that traditional models are prone to produce. Third, after the two ships pass the CPA, the model of this invention does not immediately reduce the risk to zero in any of the three scenarios. Instead, it retains a gradual residual risk output based on the distance between the ships, which naturally decays to a safe level as the distance between the two ships gradually increases.

[0089] The above results demonstrate that the three sub-models constructed in this invention exhibit consistently good performance under three typical encounter scenarios: crossover, overtaking, and head-on collision. This overcomes the risk mutation problem caused by TCPA crossing the zero point in traditional models and also compensates for the shortcomings of domain models in characterizing residual risk. The common patterns of the three curves verify the effectiveness of the core design concept of this invention: by introducing the Past state of TCPA and setting corresponding residual risk judgment rules, the model can fully characterize the risk evolution process of inland waterway vessel encounters throughout the entire cycle (from distant approach → nearest encounter point → departure), providing a more continuous and accurate risk quantification basis for inland waterway vessel risk warning and collision avoidance decisions.

[0090] Based on the above embodiments, as a preferred implementation, the dimensionless minimum encounter distance is expressed by the formula DCPA. * =DCPA / L is calculated using the formula D. * =D / L is calculated, where DCPA is the minimum encounter distance, D is the current inter-ship distance, and L is the length of the ship.

[0091] In a preferred embodiment of the present invention, the dimensionless minimum encounter distance is calculated using the formula DCPA*=DCPA / L, and the dimensionless inter-ship distance is calculated using the formula D*=D / L, where DCPA is the minimum encounter distance, D is the current inter-ship distance, and L is the length of the current ship.

[0092] In inland waterway navigation practice, there are significant differences in vessel types and sizes; the lengths of large ocean-going vessels and small inland waterway vessels can differ by several times or even tens of times. At the same absolute distance (e.g., 50 meters), for a 200-meter-long large vessel, this distance is only 0.25 times its length, representing an extremely imminent danger; however, for a 20-meter-long small vessel, this distance is 2.5 times its length, still within a relatively safe range. Therefore, directly using absolute distance units such as meters and nautical miles for risk assessment lacks a unified benchmark for comparison between vessels of different sizes, potentially leading to significant discrepancies between the assessment results and the actual situation.

[0093] To eliminate the impact of differences in ship physical dimensions on risk assessment, this invention performs dimensionless processing on the spatial dimension indicators (i.e., minimum encounter distance DCPA and current ship-to-ship distance D) before inputting them into the fuzzy inference system. Specifically, the ship's length L is selected as the dimensionless reference scale. The minimum encounter distance DCPA is divided by the ship's length L to obtain the dimensionless minimum encounter distance DCPA*; the current ship-to-ship distance D is divided by the ship's length L to obtain the dimensionless ship-to-ship distance D*. After the above conversion, DCPA and D are transformed from absolute distances into dimensionless relative values ​​relative to the ship's dimensions.

[0094] This processing provides a unified benchmark for risk assessment of vessels of different sizes. Regardless of whether the vessel is a large ocean-going vessel or a small inland waterway vessel, DCPA*=0.5 indicates a minimum encounter distance of half the vessel length, and D*=2.0 indicates a current inter-vessel distance of twice the vessel length. This allows the model to avoid biases in risk assessment caused by size differences between large and small vessels, thus significantly improving the model's applicability and versatility in encounter situations involving different vessel types on inland waterways.

[0095] Through the above technical solution, this invention inputs DCPA* and D* as dimensionless quantitative indicators of spatial proximity into the fuzzy inference system, enabling the model to adopt a unified risk judgment standard for ships of different scales, thereby improving the model's adaptability and assessment accuracy in inland waterway encounters involving multiple ship types and scales.

[0096] Based on the above embodiments, as a preferred implementation, the ship speed ratio is expressed by the formula K=V T / V O The calculation yields V. T V is the target ship's speed over land. O The speed of this vessel relative to land; the minimum meeting time TCPA is the time it takes for the two vessels to reach the nearest meeting point, and TCPA < 0 when the two vessels have passed the nearest meeting point.

[0097] In a preferred embodiment of the present invention, the ship speed ratio K is expressed by the formula K=V T / V O The calculation yields V. T V is the target ship's speed over land. O This refers to the ship's speed relative to ground. Using the ship as a reference, the speed ratio K is defined as the ratio of the target ship's speed relative to ground to the ship's speed relative to ground. The speed ratio K primarily reflects the speed difference between two ships, especially in overtaking and head-on encounters, where it significantly impacts the rate of risk escalation. In overtaking, a greater speed advantage for the following ship reduces the safety margin for the preceding ship more quickly; in head-on encounters, a higher relative speed between the two ships provides less time for judgment and maneuvering.

[0098] The minimum encounter time (TCPA) is the time it takes for two vessels to reach the nearest encounter point. TCPA < 0 when both vessels have passed the nearest encounter point. As a time-based indicator, TCPA is mainly used to determine the urgency of collision risk and reflects whether there is sufficient time for avoidance and reaction. In traditional collision hazard models, when TCPA < 0, it is generally considered that the two vessels have passed the nearest encounter point, the collision risk is eliminated, and the hazard level is directly set to zero. However, in actual inland waterway navigation, even if the two vessels have passed the nearest encounter point, if the distance between the vessels is still small, wake interference, ship suction effect, and the pressure on the pilots to make subsequent maneuvering adjustments may still exist. Therefore, in this invention, TCPA is allowed to take negative values. When TCPA < 0, it indicates that the two vessels have passed the nearest encounter point, but the model does not directly set the hazard level to zero. Instead, it still needs to combine the dimensionless distance between vessels (D*) and the dimensionless minimum encounter distance (DCPA*) to comprehensively determine whether the two vessels are still within close range, thereby retaining the ability to judge the residual risk after the vessels have passed the nearest encounter point. Taking the cross-meeting sub-model as an example, when the fuzzy linguistic variable of TCPA is in the first recently encountered point interval and the dimensionless inter-ship distance D* is still in the small value interval, the rule base retains the output of medium or high risk level; as the inter-ship distance gradually increases, the risk level decreases accordingly.

[0099] Through the above technical solution, this invention incorporates the ship speed ratio K as a quantitative indicator of motion difference into the fuzzy inference system, enabling the model to respond sensitively to changes in risk caused by speed differences. At the same time, by allowing TCPA to take negative values ​​and introducing a residual risk judgment mechanism in the Past state, the model can fully depict the entire risk evolution process of inland waterway vessels throughout the entire encounter cycle, from long-distance approach to the nearest encounter point and then to departure, overcoming the defect of traditional models that directly reduce the risk to zero after TCPA < 0.

[0100] Based on the above embodiments, as a preferred implementation, according to the current encounter situation between the vessel and the target vessel, a target sub-model corresponding to the current encounter situation is determined from a plurality of pre-constructed encounter situation sub-models, including: Based on the relative bearing and heading relationship between the ship and the target ship, the current encounter situation is identified. The relative bearing angle β describes the target ship's bearing relative to the ship and is an important basis for judging the encounter situation between the two ships; the heading relationship reflects the relative trend of the two ships' directions of motion. Together, they determine the type of the current encounter situation.

[0101] When the relative bearing angle is within the first right-hand or first left-hand interval of the cross-encounter sub-model, and the two ships' course crosses, the current encounter situation is identified as a cross-encounter situation, and the cross-encounter sub-model is determined as the target sub-model. In a cross-encounter situation, the two ships' course tends to cross, and the target ship may approach from the ship's right or left. When the target ship comes from the starboard side, the ship usually faces more direct avoidance pressure and needs to take avoidance action as early as possible; when the target ship comes from the port side, although the avoidance responsibility is different, if the two ships have already formed close pressure, the danger level will still increase accordingly. Therefore, the cross-encounter sub-model focuses on combining the relative bearing angle to determine the bearing of the approaching ship and the avoidance responsibility in the rule base design, and at the same time integrates indicators such as DCPA*, TCPA, D*, and K for risk reasoning.

[0102] When the relative bearing is within the second right rear or second left rear interval of the overtaking sub-model, and both ships are heading in the same direction with one ship having a higher speed than the other, the current encounter situation is identified as an overtaking situation, and the overtaking sub-model is determined as the target sub-model. In an overtaking situation, one ship approaches from behind another ship and intends to overtake it. The two ships are heading in roughly the same direction, and the speed of the rear ship is greater than that of the front ship. The relative bearing being within the stern direction interval is the primary condition for judging an overtaking situation. An overtaking relationship can only be formed when the target ship is within a certain range behind the ship. The same heading and the existence of a speed difference further confirm that the two ships are traveling in the same direction and the rear ship is approaching the front ship. Therefore, the overtaking sub-model focuses on combining the ship speed ratio K to judge the degree of speed pressure in the rule base design, and comprehensively considers indicators such as DCPA*, TCPA, D*, and β to judge the risk evolution during the overtaking process.

[0103] When the relative bearing is within the third right-forward or third left-forward interval of the encounter sub-model, and the two ships are heading in opposite or nearly opposite directions, the current encounter situation is identified as an encounter situation, and the encounter sub-model is determined as the target sub-model. In an encounter situation, the two ships approach each other head-on in opposite or nearly opposite directions, with the relative bearing usually near the bow direction. The target ship is located to the right or left-forward of the current ship, indicating that the two ships are gradually approaching from opposite directions. Opposite or nearly opposite headings confirm a head-on collision in the two ships' directions of motion. During an encounter, the relative speed of the two ships is relatively high, and the time available for judgment and maneuvering by the pilot is short, so the risk often develops more quickly. Therefore, the encounter sub-model places greater emphasis on the dominant role of TCPA and D* in the hazard change in its rule base design. When the TCPA is short and the distance is close, the hazard level increases significantly.

[0104] Through the above technical solution, this invention can automatically and accurately identify whether the current encounter situation is a cross encounter, overtaking, or head-on encounter based on the relative bearing and heading relationship between the vessel and the target vessel, and adaptively select the corresponding target sub-model for reasoning. This enables the model to use differentiated membership functions and fuzzy rule bases for risk assessment under different encounter situations, avoiding the problem of risk judgment distortion in traditional unified models under different encounter situations, and improving the situation adaptability and assessment accuracy of inland waterway vessel collision risk quantification.

[0105] Based on the above embodiments, as a preferred implementation, in the cross-meeting sub-model, there is an overlapping area between the first right interval and the first stern direction interval, and there is an overlapping area between the first stern direction interval and the first left interval.

[0106] In the cross-encounter sub-model, the fuzzy linguistic variable intervals for the relative bearing angle β include the first right-hand interval, the first stern-hand interval, and the first left-hand interval. In a cross-encounter situation, when the target vessel comes from the starboard side, the vessel typically faces more direct avoidance pressure; when the target vessel comes from the port side or stern-hand, the avoidance responsibility and risk assessment logic differs. Therefore, the interval division of β directly relates to the risk level determination result in the rule base.

[0107] To prevent abrupt changes in risk values ​​due to relative azimuth angles falling precisely near interval boundaries—for example, when the target ship's azimuth changes from 89° to 91°—if there is no overlap between intervals, the model might jump from the "right-hand interval" to the "stern-direction interval" in a very short time, causing an unreasonable jump or drop in collision risk. This invention incorporates overlapping regions between adjacent fuzzy linguistic variable intervals. Specifically, there is an overlapping region between the first right-hand interval and the first stern-direction interval, and between the first stern-direction interval and the first left-hand interval. When the target ship's bearing gradually transitions from starboard to stern, or from stern to port, due to the overlap between adjacent intervals, the input variable simultaneously belongs to two adjacent intervals with different membership degrees during the transition. The model output will not abruptly change simply because the azimuth angle crosses a boundary; instead, it can achieve a smooth transition between adjacent states.

[0108] Taking the transition of the target ship's bearing from the right-hand zone to the stern-direction zone as an example: when the bearing is in the core area of ​​the first right-hand zone, its membership degree to the "right-hand zone" is 1, and its membership degree to the "stern-direction zone" is 0; when the bearing enters the overlapping area of ​​the two zones, its membership degree to the "right-hand zone" gradually decreases, while its membership degree to the "stern-direction zone" gradually increases; when the bearing completely enters the core area of ​​the first stern-direction zone, its membership degree to the "right-hand zone" is 0, and its membership degree to the "stern-direction zone" is 1. During this process, the membership degrees of the two zones exhibit a continuous, inverse relationship, ensuring a smooth and continuous collision risk output from the model, without any abrupt changes caused by the angle crossing boundaries.

[0109] Through the above technical solution, this invention features overlapping regions between adjacent intervals of relative azimuth angles in the cross-encounter sub-model. This allows the model to maintain the smoothness and continuity of risk output even when the target ship's azimuth changes continuously, avoiding abrupt changes in risk values ​​due to minor parameter variations. This makes the changes in collision risk more closely reflect the driver's continuous perception of the risk situation during actual navigation. This design is also applicable to the overlapping settings between adjacent intervals of relative azimuth angles in the overtaking and encounter sub-models.

[0110] Based on the above embodiments, as a preferred implementation, in the overtaking sub-model, there are overlapping areas between the second right front interval and the second right lateral direction region interval, between the second right lateral direction region interval and the second right rear interval, between the second right rear interval and the second left rear interval, between the second left rear interval and the second left lateral direction region interval, and between the second left lateral direction region interval and the second left front interval.

[0111] In the overtaking sub-model, the fuzzy linguistic variable interval for the relative bearing angle β includes six regions: the second right forward region (Front_R), the second right transverse region (Side_R), the second right rear region (Stern_R), the second left rear region (Stern_L), the second left transverse region (Side_L), and the second left forward region (Front_L). This division is closely related to the actual changes in the overtaking situation. Overtaking is not an instantaneous process; the target vessel may gradually approach from the stern, then enter the transverse position, and finally gradually reach the bow. The bearing changes themselves reflect the progress of the overtaking process. Using these six regions, the model can distinguish between different stages such as tail-following, passing alongside, and clearing the way.

[0112] To prevent abrupt changes in risk values ​​due to relative azimuth angles falling near the boundaries of adjacent intervals, this invention incorporates overlapping regions between adjacent intervals of relative azimuth angles in the overtaking sub-model. Specifically, overlapping regions exist between the second right forward interval and the second right lateral direction interval, between the second right lateral direction interval and the second right rear interval, between the second right rear interval and the second left rear interval, between the second left rear interval and the second left lateral direction interval, and between the second left lateral direction interval and the second left forward interval.

[0113] Taking the approach of a target vessel from its stern towards its beam position during an overtaking maneuver as an example: when the target vessel's relative bearing is within the core region of the second right aft interval, its membership degree to the "right aft interval" is 1, and its membership degree to the "right beam region interval" is 0. As the target vessel approaches and enters the overlapping region of the two intervals, its membership degree to the "right aft interval" gradually decreases, while its membership degree to the "right beam region interval" gradually increases. When the target vessel fully enters the core region of the second right beam region interval, its membership degree to the "right aft interval" is 0, and its membership degree to the "right beam region interval" is 1. Throughout this process, the membership degrees of the two adjacent intervals exhibit a continuous, inverse relationship, ensuring that the collision risk output by the model remains continuous and smooth throughout the entire overtaking process, from the target vessel gradually approaching from its stern position, to entering the beam position for close-range maneuvering, and then gradually passing and clearing the way. This prevents abrupt changes due to the bearing crossing interval boundaries.

[0114] Furthermore, an overlapping area is also set between the second right rear section and the second left rear section. In an inland waterway overtaking situation, the target vessel may approach from the right rear or left rear of the vessel. The overlapping area between the two sections ensures that when the target vessel transitions from one side to the other in the direction directly aft of the vessel (for example, from right rear to directly aft and then to left rear), the model output will not change abruptly because the angle just crosses the 180° boundary, thus keeping the judgment of the overtaking situation stable when the bearing changes continuously.

[0115] Through the above technical solution, this invention features overlapping regions between the six adjacent intervals of the relative azimuth angle in the overtaking sub-model, enabling the model to maintain the smoothness and continuity of risk output even when the target vessel's azimuth changes continuously. This design is particularly suitable for overtaking situations on inland waterways, where the overtaking duration is relatively long and the target vessel's azimuth gradually changes from stern to bow, with the azimuth angle spanning multiple interval boundaries. By overlapping the adjacent intervals, the collision risk output by the model can continuously reflect the risk evolution throughout the entire overtaking process, from tail-to-tail approach to parallel passage and finally to clearing the way, avoiding abrupt changes in risk values ​​due to minor parameter variations. This makes the risk assessment more closely reflect the driver's continuous perception of the overtaking situation during actual navigation.

[0116] Based on the above embodiments, as a preferred implementation, in the encounter sub-model, there are overlapping areas between the third right front interval and the third right side interval, between the third right side interval and the third left side interval, and between the third left side interval and the third left front interval.

[0117] In the encounter sub-model, the fuzzy linguistic variable intervals for the relative azimuth angle β include four regions: the third right-front interval (Front_R), the third right-side interval (Side_R), the third left-side interval (Side_L), and the third left-front interval (Front_L). Unlike the cross-encounter sub-model, which focuses on distinguishing the bearing of the approaching vessel and the responsibility for avoidance, the relative azimuth angle division in the encounter sub-model focuses on determining whether the target vessel is near its bow and whether the two vessels are forming a head-on approach. For encounters on curves, the target vessel may gradually appear from the left or right of its bow; therefore, each boundary interval needs to retain a certain degree of overlap to avoid sudden changes in risk assessment caused by small changes in azimuth angle.

[0118] To prevent sudden changes in risk values ​​due to relative azimuth angles being near the boundaries of adjacent intervals, for example, when the target ship's azimuth angle gradually transitions from the right-hand side to the right-hand side, if there is no overlap between intervals, the model may jump from the "right-hand front interval" to the "right-hand side interval" in a very short time, resulting in an unreasonable jump or drop in collision risk.

[0119] This invention features overlapping regions between adjacent azimuth intervals in the encounter model. Specifically, there is an overlapping region between the third right-forward interval and the third right-side interval, between the third right-side interval and the third left-side interval, and between the third left-side interval and the third left-forward interval. When the target ship's azimuth gradually transitions from right-forward to right-side, or from right-side to left-side, or from left-side to left-forward, due to the overlap between adjacent intervals, the input variables will simultaneously belong to two adjacent intervals with different membership degrees during the transition. The model output will not abruptly change simply because the azimuth angle crosses a boundary; instead, it can achieve a smooth transition between adjacent states.

[0120] Taking the transition of the target ship's bearing from the right-forward interval to the right-right interval as an example: when the bearing is in the core area of ​​the third right-forward interval, its membership degree to the "right-forward interval" is 1, and its membership degree to the "right-right interval" is 0; when the bearing enters the overlapping area of ​​the two intervals, its membership degree to the "right-forward interval" gradually decreases, while its membership degree to the "right-right interval" gradually increases; when the bearing completely enters the core area of ​​the third right-right interval, its membership degree to the "right-forward interval" is 0, and its membership degree to the "right-right interval" is 1. During this process, the membership degrees of the two intervals exhibit a continuous, inverse relationship, ensuring that the collision risk output by the model is continuous and smooth, without any abrupt changes caused by the angle crossing boundaries.

[0121] In a head-on situation, the third right forward interval and the third left forward interval correspond to the target vessel being on the right and left side of the bow, respectively, indicating a direct approach. When the target vessel moves from the right side of the bow to the right (the approach gradually weakens), or moves from the right side to directly forward and then to the left (the course changes from port or starboard), the overlapping area between adjacent intervals ensures that the model output maintains a smooth transition during continuous changes in azimuth, without causing jumps in risk values ​​due to the angle crossing boundaries.

[0122] Through the above technical solution, this invention features overlapping regions between adjacent intervals of the relative azimuth angle in the encounter sub-model, enabling the model to maintain the smoothness and continuity of risk output even when the target vessel's azimuth changes continuously. This design is particularly suitable for encounter situations in winding inland waterways, where the target vessel may gradually appear from the left or right of the bow, with the azimuth angle spanning multiple interval boundaries. By overlapping adjacent intervals, the collision risk output by the model can continuously reflect the entire risk evolution process from the head-on approach phase to the completion of the meeting, avoiding abrupt changes in risk values ​​due to minor parameter variations. This makes the risk assessment more closely reflect the driver's continuous perception of the encounter situation on a winding road during actual navigation.

[0123] In a preferred embodiment of the present invention, the reasoning result obtained after fuzzy rule reasoning is a fuzzy output set formed by the aggregation of multiple rule conclusions. To further obtain an accurate collision risk value for navigation decision-making, the present invention uses the centroid method to defuzzify the fuzzy output set. The centroid method calculates the abscissa value corresponding to the centroid of the region enclosed by the membership function curve of the fuzzy output set and the horizontal axis, which is used as the accurate collision risk output value after defuzzification. The centroid method is simple to calculate and has good continuity, which can avoid risk value jumps caused by improper selection of defuzzification method, so that the output result transitions smoothly and matches the continuous perception of the degree of danger by the actual driver. Through the above defuzzification process, each sub-model finally outputs an accurate collision risk value between 0 and 1, where a higher value indicates a greater collision risk in the current situation.

[0124] Secondly, embodiments of the present invention provide a collision risk quantification system for inland waterway vessels in multi-encounter situations, based on the collision risk quantification methods for inland waterway vessels in the above embodiments, such as... Figure 5 As shown, the system 500 includes: The navigation status parameter acquisition module 510 is used to acquire the navigation status parameters of the ship and the target ship, including the ship's position, speed, heading and length.

[0125] The evaluation index calculation module 520 is connected to the navigation state parameter acquisition module and is used to calculate the collision risk evaluation index between the ship and the target ship based on the navigation state parameters. The collision risk evaluation index includes minimum encounter distance, minimum encounter time, current ship-to-ship distance, relative bearing angle and ship speed ratio.

[0126] The dimensionless processing module 530, connected to the evaluation index calculation module, is used to perform dimensionless processing on the minimum encounter distance and the current inter-ship distance using the ship's length as the dimensionless reference scale, so as to obtain the dimensionless minimum encounter distance and the dimensionless inter-ship distance.

[0127] The meeting situation sub-model storage module 540 is used to store multiple pre-built meeting situation sub-models, including cross-meeting sub-models, overtaking sub-models, and face-to-face meeting sub-models; the multiple meeting situation sub-models have different membership function threshold parameter configurations and different fuzzy rule bases; when the minimum meeting time is negative, it indicates that the two ships have passed the nearest meeting point.

[0128] The encounter situation identification and sub-model selection module 550 is connected to the dimensionless processing module and the encounter situation sub-model storage module, and is used to determine the target sub-model corresponding to the current encounter situation from the encounter situation sub-model storage module based on the current encounter situation between the ship and the target ship.

[0129] The fuzzy reasoning module 560 is connected to the dimensionless processing module and the encounter situation recognition and sub-model selection module. It is used to input the dimensionless minimum encounter distance, the minimum encounter time, the dimensionless inter-ship distance, the relative azimuth angle and the ship speed ratio into the target sub-model, and output the collision risk degree based on fuzzy logic reasoning.

[0130] Where there is no conflict, the technical features in this embodiment and implementation scheme can be combined arbitrarily.

Claims

1. A method for quantifying collision risk in multi-encounter situations of inland waterway vessels, characterized in that, The method includes: Obtain the navigation status parameters of the ship and the target ship, including the ship's position, speed, heading and length; Based on the navigation status parameters, calculate the collision risk assessment index between the ship and the target ship. The collision risk assessment index includes the minimum encounter distance, the minimum encounter time, the current ship-to-ship distance, the relative bearing angle, and the ship speed ratio. Using the ship's length as a dimensionless reference scale, the minimum meeting distance and the current ship-to-ship distance are processed to obtain the dimensionless minimum meeting distance and the dimensionless ship-to-ship distance. Based on the current encounter situation between the vessel and the target vessel, a target sub-model corresponding to the current encounter situation is determined from multiple pre-constructed encounter situation sub-models. These multiple encounter situation sub-models include a cross-encounter sub-model, an overtaking sub-model, and a face-to-face sub-model. Each encounter situation sub-model uses a Mamdani-type IF-THEN rule to construct an independent fuzzy rule base, and each of the multiple encounter situation sub-models has a unique membership function threshold parameter configuration. Each independent fuzzy rule base is also unique. When the minimum encounter time is negative, it indicates that the two vessels have passed the nearest encounter point. The dimensionless minimum encounter distance, the minimum encounter time, the dimensionless inter-ship distance, the relative azimuth angle, and the ship speed ratio are input into the target sub-model, and the target sub-model outputs the collision risk level based on fuzzy logic reasoning.

2. The method for quantifying collision risk in multi-encounter situations of inland waterway vessels according to claim 1, characterized in that, Multiple encounter scenario sub-models use the dimensionless minimum encounter distance, the minimum encounter time, the dimensionless inter-ship distance, the relative azimuth angle, and the ship speed ratio as input variables, and the collision risk as the output variable. Each of the aforementioned encounter scenario sub-models employs a trapezoidal membership function to characterize the membership degree of each input variable under different risk levels. Each input variable is divided into multiple fuzzy linguistic variable intervals, with overlapping regions between adjacent intervals. In the cross-encounter sub-model, the fuzzy linguistic variable interval for the minimum encounter time includes the interval of the first nearest encounter point that has passed, the interval of the first time urgency state, the interval of the first medium urgency state, and the interval of the first time sufficiency state; the fuzzy linguistic variable interval for the relative azimuth angle includes the interval of the first right side, the interval of the first stern direction, and the interval of the first left side. In the overtaking sub-model, the fuzzy linguistic variable interval for the minimum encounter time includes the interval of the second nearest encounter point that has been passed, the interval of the second time urgency state, the interval of the second medium urgency state, and the interval of the second time sufficiency state; the fuzzy linguistic variable interval for the relative azimuth angle includes the interval of the second right front, the interval of the second right lateral direction, the interval of the second right rear, the interval of the second left rear, the interval of the second left lateral direction, and the interval of the second left front. In the encounter sub-model, the fuzzy linguistic variable interval for the minimum encounter time includes the interval for the third most recent encounter point that has passed, the interval for the third time-urgent state, the interval for the third medium-urgent state, and the interval for the third time-sufficient state; the fuzzy linguistic variable interval for the relative azimuth angle includes the interval for the third right front, the interval for the third right side, the interval for the third left side, and the interval for the third left front.

3. The method for quantifying collision risk in multi-encounter situations of inland waterway vessels according to claim 1, characterized in that, The dimensionless minimum encounter distance is expressed by the formula DCPA. * =DCPA / L is calculated using the formula D. * =D / L is calculated, where DCPA is the minimum encounter distance, D is the current inter-ship distance, and L is the length of the ship.

4. The method for quantifying collision risk in multi-encounter situations of inland waterway vessels according to claim 1, characterized in that, The ship speed ratio is expressed by the formula K=V T / V O The calculation yields V. T V is the target ship's speed over land. O The speed of this vessel relative to land; the minimum meeting time TCPA is the time it takes for the two vessels to reach the nearest meeting point, and TCPA < 0 when the two vessels have passed the nearest meeting point.

5. The method for quantifying collision risk in multi-encounter situations of inland waterway vessels according to claim 2, characterized in that, Based on the current encounter situation between the vessel and the target vessel, the target sub-model corresponding to the current encounter situation is determined from multiple pre-constructed encounter situation sub-models, including: Identify the current encounter situation based on the relative bearing and course relationship between the vessel and the target vessel; When the relative azimuth angle is in the first right interval or the first left interval of the cross encounter sub-model, and the two ships' headings intersect, the current encounter situation is identified as a cross encounter situation, and the cross encounter sub-model is determined as the target sub-model. When the relative azimuth angle is in the second right rear interval or the second left rear interval in the overtaking sub-model, and the two ships are heading in the same direction and one ship's speed is greater than the other ship's speed, the current encounter situation is identified as an overtaking situation, and the overtaking sub-model is determined as the target sub-model. When the relative azimuth angle is within the third right forward interval or the third left forward interval in the encounter sub-model, and the two ships are heading in opposite or nearly opposite directions, the current encounter situation is identified as an encounter situation, and the encounter sub-model is determined as the target sub-model.

6. The method for quantifying collision risk in multi-encounter situations of inland waterway vessels according to claim 2, characterized in that, In the intersection encounter sub-model, there is an overlapping area between the first right interval and the first stern direction interval, and there is an overlapping area between the first stern direction interval and the first left interval.

7. The method for quantifying collision risk in multi-encounter situations of inland waterway vessels according to claim 2, characterized in that, In the overtaking sub-model, there are overlapping areas between the second right front interval and the second right horizontal direction region interval, between the second right horizontal direction region interval and the second right rear interval, between the second right rear interval and the second left rear interval, between the second left rear interval and the second left horizontal direction region interval, and between the second left horizontal direction region interval and the second left front interval.

8. The method for quantifying collision risk in multi-encounter situations of inland waterway vessels according to claim 2, characterized in that, In the aforementioned encounter sub-model, there are overlapping areas between the third right front interval and the third right side interval, between the third right side interval and the third left side interval, and between the third left side interval and the third left front interval.

9. The method for quantifying collision risk in multi-encounter situations of inland waterway vessels according to claim 2, characterized in that, The trapezoidal membership function is calculated using the following formula: Where x is the value of the input variable, and a, b, c, and d are the threshold parameters of each input variable in the corresponding fuzzy linguistic variable range; The fuzzy rule bases of the cross-encounter sub-model, the overtaking sub-model, and the encounter sub-model all use four risk levels—low risk, medium risk, high risk, and extremely high risk—as output fuzzy language variables; the fuzzy sets corresponding to each risk level are defuzzified and converted into collision risk index (CRI) values ​​in the range of 0 to 1. In the fuzzy rule base of the cross-encounter sub-model, when the first right interval, the first time urgency interval, the minimum value interval of the dimensionless minimum encounter distance, and the minimum value interval of the dimensionless inter-ship distance are all met simultaneously, it is determined to be the extremely high risk; when the first interval of the nearest encounter point that has passed and the minimum value interval of the dimensionless inter-ship distance are both met simultaneously, the output of medium risk or high risk is retained. In the fuzzy rule base of the overtaking sub-model, when the minimum interval of the dimensionless minimum encounter distance, the minimum interval of the dimensionless inter-ship distance, and the second right transverse region interval or the second left transverse region interval are all true, it is determined to be the extremely high risk; when the second interval of the nearest passing point and the minimum interval of the dimensionless inter-ship distance or the minimum interval of the dimensionless minimum encounter distance are all true, the output of medium risk or high risk is retained. In the fuzzy rule base of the encounter sub-model, when the time urgency interval of the minimum encounter time and the small value interval of the dimensionless inter-ship distance are both met, it is determined to be the extremely high risk; when the interval of the third nearest encounter point that has passed and the small value interval of the dimensionless inter-ship distance are both met, the medium risk output is retained until the large value interval of the dimensionless inter-ship distance is met, and then it is reduced to low risk.

10. A system for quantifying collision risk in multi-encounter situations of inland waterway vessels, characterized in that, The system includes: The navigation status parameter acquisition module is used to acquire the navigation status parameters of the ship and the target ship, including the ship's position, speed, heading and length. The evaluation index calculation module is connected to the navigation state parameter acquisition module and is used to calculate the collision risk evaluation index between the ship and the target ship based on the navigation state parameters. The collision risk evaluation index includes minimum encounter distance, minimum encounter time, current ship-to-ship distance, relative bearing angle and ship speed ratio. The dimensionless processing module, connected to the evaluation index calculation module, is used to perform dimensionless processing on the minimum encounter distance and the current inter-ship distance using the ship's length as the dimensionless benchmark scale, to obtain the dimensionless minimum encounter distance and the dimensionless inter-ship distance. The meeting situation sub-model storage module is used to store multiple pre-built meeting situation sub-models, including cross-meeting sub-models, overtaking sub-models, and face-to-face meeting sub-models. Each meeting situation sub-model is constructed with an independent fuzzy rule base using the Mamdani-type IF-THEN rule format, and the multiple meeting situation sub-models have different membership function threshold parameter configurations. The independent fuzzy rule bases are also different from each other. When the minimum meeting time is negative, it indicates that the two ships have passed the nearest meeting point. The encounter situation identification and sub-model selection module is connected to the dimensionless processing module and the encounter situation sub-model storage module. It is used to determine the target sub-model corresponding to the current encounter situation from the encounter situation sub-model storage module based on the current encounter situation between the ship and the target ship. The fuzzy reasoning module, connected to the dimensionless processing module and the encounter situation recognition and sub-model selection module, is used to input the dimensionless minimum encounter distance, the minimum encounter time, the dimensionless inter-ship distance, the relative azimuth angle and the ship speed ratio into the target sub-model, and output the collision risk level based on fuzzy logic reasoning.

Images