A ship collision warning system based on collision avoidance rules

Abstract

This invention discloses a ship collision warning system based on collision avoidance rules, belonging to the field of ship navigation safety technology. Addressing the shortcomings of existing ship collision warning systems, such as low compatibility with international maritime collision avoidance rules, insufficient risk calculation accuracy, and lack of support for compliant collision avoidance decisions, this system consists of a multi-source data acquisition and preprocessing module, an encounter scenario identification module, a collision avoidance rule inference engine, a collision risk quantification assessment module, a graded warning and collision avoidance decision generation module, and a human-computer interaction module. This invention incorporates a collision avoidance rule knowledge base, automatically identifying typical encounter scenarios such as face-to-face encounters, intersections, overtaking, and poor visibility, matching corresponding rule clauses, and clarifying the ship's avoidance responsibilities and action constraints. It integrates multi-source data from AIS, radar, electronic charts, and meteorological and hydrological data, combines ship maneuvering parameters to correct the collision risk calculation model, outputs graded warnings, generates compliant quantified collision avoidance schemes, and completes trajectory simulation verification, significantly improving warning accuracy and collision avoidance decision compliance, effectively ensuring ship navigation safety in complex navigation scenarios.

Description

Technical Field

[0001] This invention relates to the field of ship navigation safety technology, and in particular to a ship collision warning system based on collision avoidance rules. Background Technology

[0002] Ship collisions are among the most frequent and serious types of maritime safety accidents. Collisions not only cause damage to ships, cargo loss, and personal injury, but can also trigger secondary disasters such as oil spills, fires, and marine pollution, resulting in huge economic losses and severe social impacts. Most existing ship collision warning systems rely on Distance to Closest Approach (DCPA) and Time to Closest Approach (TCPA) as core indicators, triggering warnings by setting fixed thresholds. However, this approach has significant shortcomings in practical applications. First, the system fails to deeply integrate the specific provisions of the 1972 International Regulations for Preventing Collisions at Sea (COLREGs), relying solely on distance and time for risk assessment. This fails to differentiate between different encounter scenarios such as head-on collisions, crossings, and overtaking, and also fails to clarify the yielding responsibilities between vessels. The resulting warnings are disconnected from navigation rules, easily leading to misjudgments and operational errors by navigators. According to maritime statistics, over 80% of ship collisions are directly related to human factors, with navigators' misunderstandings of collision avoidance rules, misjudgments of encounter situations, and improper timing of yielding being the core contributing factors. The existing system cannot compensate for these human cognitive shortcomings at the decision-making level.

[0003] Secondly, the data sources are limited. Most systems rely solely on AIS data for calculations, failing to integrate radar detection, electronic charts, meteorological and hydrological data, and the ship's own maneuvering parameters. Under complex sea conditions such as wind and current interference and poor visibility, the calculation errors of encounter parameters are large, resulting in high rates of missed and false alarms in early warnings. In complex navigation environments such as narrow waterways, lane separation systems, and areas with dense fishing vessels, scenarios involving multiple vessels encountering each other frequently occur. Most existing systems only support collision risk calculations for single vessels, failing to perform global risk ranking and comprehensive avoidance decisions for multi-target encounters. This can easily lead to avoidance conflicts where one aspect is overlooked, thereby exacerbating navigation risks.

[0004] Third, the functionality is limited to risk warnings and does not provide quantitative collision avoidance decision-making schemes that comply with collision avoidance rules. Drivers must make their own judgments on avoidance actions in emergency situations, resulting in low emergency response efficiency and the risk of non-compliant actions due to human decision-making. Existing solutions that introduce collision avoidance rules are generally poorly adapted, merely converting rule clauses into simple conditional judgments and failing to cover flexible clauses such as "small-angle intersection boundaries" and "continuity of liability for overtaking." The rule-based reasoning results deviate from actual navigation practices. Furthermore, the output collision avoidance suggestions are mostly qualitative descriptions, without considering the vessel's load, draft, and main engine performance to verify feasibility, thus limiting their practical value.

[0005] Fourth, it is impossible to pre-simulate and verify collision avoidance actions. The avoidance plan formulated by the pilot may cause a secondary collision or violate the collision avoidance rules. There is a lack of a closed-loop verification mechanism for the effectiveness of avoidance, making it difficult to assess whether the avoidance actions comply with the "early, large, wide, and clear" collision avoidance action principles, and making it difficult to meet the requirements of high-level navigation safety.

[0006] Therefore, developing a ship collision warning system that deeply conforms to collision avoidance rules, integrates multi-source data, and has both risk warning and compliant collision avoidance decision-making capabilities is a technical problem that urgently needs to be solved in this field. Summary of the Invention

[0007] To address the shortcomings of existing technologies, the present invention aims to provide a ship collision warning system based on collision avoidance rules. Through a closed-loop design that integrates intelligent encounter scenario recognition, dynamic reasoning of collision avoidance rules, risk assessment corrected by multi-source data, and graded warning and compliant collision avoidance decision deduction, the system solves the problems of low compatibility between existing warning systems and collision avoidance rules, insufficient accuracy of risk calculation, and lack of support for compliant decision-making. This comprehensively improves the accuracy, compliance, and safety of ship navigation warnings.

[0008] To achieve the above objectives, the present invention provides the following technical solution: A ship collision warning system based on collision avoidance rules includes: The multi-source data acquisition and preprocessing module is used to collect navigation perception data, environmental data and ship handling parameters of the ship and the target ship, perform spatiotemporal registration, outlier data removal and data completion, and output a standardized navigation dataset. The encounter scene recognition module is used to calculate the bearing, distance, relative speed and heading difference of the target ship relative to the ship based on the standardized navigation dataset, and automatically identify the encounter scene type according to the maritime collision avoidance rules; The collision avoidance rule reasoning engine has a built-in collision avoidance rule knowledge base, which is used to map the identified encounter scenario types to match the corresponding collision avoidance rule clauses, and output the division of avoidance responsibility and avoidance action constraints between the ship and the target ship. The collision risk quantification assessment module is used to combine the encounter scenario type and avoidance responsibility, calculate the corrected nearest encounter distance and nearest encounter time, introduce ship handling parameters and environmental disturbances to construct a risk assessment model, and output the collision risk level. The graded early warning and collision avoidance decision generation module is used to trigger early warning signals of the corresponding level according to the collision risk level, generate quantitative collision avoidance action plans based on avoidance action constraints, and deduce the navigation trajectory after the collision avoidance action to verify the safety of the plan. The human-computer interaction module is used to display the encounter situation, collision risk level, early warning information and collision avoidance action plan in real time, and to receive manual intervention instructions.

[0009] Furthermore, the navigation perception data collected by the multi-source data acquisition and preprocessing module includes AIS data, radar detection data, and electronic chart data; environmental data includes wind speed, wind direction, current speed, current direction, and visibility data; and ship maneuvering parameters include the ship's turning radius, braking distance, and maneuvering response delay parameters. The preprocessing module uses a spatiotemporal interpolation algorithm to achieve unified registration of timestamps and spatial coordinates of the multi-source data.

[0010] Furthermore, the encounter scene recognition module identifies encounter scene types including face-to-face encounters, cross-encounter encounters, overtaking encounters, and poor visibility scenarios; among which, the condition for determining a face-to-face encounter is that the difference in heading between the target ship and the ship is in the range of 165°~195°, and the condition for determining an overtaking encounter is that the target ship is at least 22.5° aft of the ship's main beam and the ship's speed is greater than that of the target ship.

[0011] Furthermore, the collision avoidance rule knowledge base is built on the International Regulations for Preventing Collisions at Sea, 1972, and uses a production rule representation to store the yielding responsibilities, code of conduct, and prohibited behaviors corresponding to each encounter scenario; the collision avoidance rule reasoning engine adopts a forward reasoning mechanism, which triggers the corresponding rule clauses based on the input encounter scenario and the ship's situation.

[0012] Furthermore, the correction process of the collision risk quantification assessment module is as follows: the relative motion trajectory is corrected based on the offset of the ship's trajectory by the wind and air pressure, and the calculation results of the encounter distance and encounter time are corrected in combination with the ship's turning and braking performance parameters; the risk assessment model adopts the fuzzy comprehensive evaluation method, and takes the corrected nearest encounter distance, nearest encounter time and avoidance responsibility weight as inputs to output a four-level collision risk level.

[0013] Furthermore, the warning levels of the graded warning and collision avoidance decision generation module include low-risk warning, medium-risk warning, high-risk warning and emergency risk warning, which correspond to warning methods such as sound and light prompts, text prompts, forced pop-ups and sound and light alarm combinations, respectively.

[0014] Furthermore, the quantified collision avoidance action plan includes the turning direction and angle, speed adjustment amount, and optimal timing of action; the trajectory simulation uses a ship motion mathematical model to simulate the navigation trajectory after the collision avoidance action is executed, and verifies whether the collision risk under the new trajectory has been reduced to below the safety threshold.

[0015] Furthermore, the human-computer interaction module is integrated into the ship's bridge display terminal, supporting the overlay display of the target ship's situation, risk areas and recommended collision avoidance routes on electronic charts, and allowing the driver to manually set warning thresholds and rule adaptation parameters.

[0016] Compared with the prior art, the present invention has the following beneficial effects: 1. Deep integration of rules and strong compliance of early warning: Breaking through the limitations of traditional threshold-based early warning, the international maritime collision avoidance rules are transformed into calculable logical constraints through automatic identification of encounter scenarios and rule reasoning engine, clarifying the responsibility and action boundaries of ships to avoid collisions. Both early warning and decision-making strictly comply with navigation rules, reducing the risk of human misjudgment of rules from the root.

[0017] 2. Multi-source data correction for high assessment accuracy: By integrating multi-source data from AIS, radar, electronic charts, and meteorological and hydrological sources, and introducing a calculation model for encounter parameters that corrects for ship maneuverability parameters and wind, current, and pressure interference, the accuracy of collision risk assessment under complex sea conditions and poor visibility is significantly improved, reducing missed and false alarms.

[0018] 3. Full-link closed-loop design with strong decision support capabilities: Constructing a complete closed loop of "situational awareness - scene recognition - rule reasoning - risk assessment - graded early warning - decision generation - trajectory verification", it not only outputs risk warnings, but also automatically generates compliant quantitative collision avoidance solutions and pre-verifies safety, which greatly improves the driver's emergency response efficiency and collision avoidance operation reliability.

[0019] 4. Modular and scalable with wide adaptability: The system adopts a modular architecture, the rule knowledge base can be updated according to the rules of different navigation areas, and the operation parameters can be adapted to different ship types. It can be widely used in bridge auxiliary decision-making systems for various types of ships such as merchant ships, government ships, and engineering ships. Attached Figure Description

[0020] Figure 1 is a block diagram of the overall architecture of the ship collision warning system based on collision avoidance rules according to the present invention; Figure 2 is a flowchart of the encounter scene recognition and collision avoidance rule reasoning of the present invention; Figure 3 is a flowchart of the collision risk assessment and collision avoidance decision generation process of the present invention. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0022] The present invention proposes a ship collision warning system based on collision avoidance rules, comprising the following modules: Multi-source data acquisition and preprocessing module: The system collects various types of navigation perception data, environmental data, and ship handling parameters from both the ship and the target vessel. It performs spatiotemporal registration, outlier removal, and missing data completion on the multi-source heterogeneous data to generate a standardized navigation dataset with unified spatiotemporal characteristics. The navigation perception data includes AIS dynamic data, radar echo data, and electronic chart static data; the environmental data includes wind speed, wind direction, current speed, current direction, and visibility data; and the ship handling parameters include the ship's turning radius, braking distance, rudder response delay, and main engine acceleration characteristics. Meeting Scene Intelligent Recognition Module: Based on standardized navigation datasets, the system calculates the real-time bearing, distance, relative velocity vector, and heading difference of the target vessel relative to itself. According to the scenario definitions of maritime collision avoidance rules, the system automatically identifies the types of encounter scenarios, including face-to-face situations, cross-encounter situations, overtaking situations, and scenarios with poor visibility. Collision avoidance rule reasoning engine: It has a built-in collision avoidance rule knowledge base based on COLREGs, and uses production rule representation to store the yield responsibility division, avoidance action criteria and prohibited behaviors corresponding to each encounter scenario; it adopts a forward reasoning mechanism, takes the identified encounter scenario and real-time ship status as input, matches and triggers the corresponding rule clauses, and outputs the qualitative result of the ship's avoidance responsibility and action constraints. Collision risk quantitative assessment module: Combining the encounter scenario type and avoidance responsibility weight, the basic DCPA and TCPA are calculated based on the principle of relative motion; wind, current, pressure, track offset, and ship maneuvering parameters are introduced to correct the basic encounter parameters, and a multi-dimensional risk assessment model is constructed; the fuzzy comprehensive evaluation method is used to calculate the collision risk value, and four risk levels—low risk, medium risk, high risk, and emergency risk—are classified and output; Tiered early warning and collision avoidance decision generation module: Based on the collision risk level, a warning signal of the corresponding level is triggered; at the same time, based on the action constraints output by the collision avoidance rules, with the goal of minimizing collision risk and meeting the rule requirements, a quantitative collision avoidance action plan is generated, including the turning direction and angle, speed adjustment amount and the best action timing; the ship motion mathematical model is called to deduce the navigation trajectory after the collision avoidance action is executed, and it is verified whether the collision risk under the new trajectory has dropped below the safety threshold. If it does not meet the requirement, the collision avoidance plan is iteratively optimized. Human-computer interaction module: Integrated into the ship's bridge display terminal, it overlays the target ship's situation, encounter scenarios, risk areas, and recommended collision avoidance routes onto the electronic chart; it displays risk levels, warning information, and collision avoidance action plans in real time; and it supports the driver to manually adjust warning thresholds, rule adaptation parameters, and manually intervene in the generation logic of collision avoidance plans.

[0023] The present invention will now be described in further detail with reference to the accompanying drawings and specific embodiments.

[0024] A ship collision early warning system based on collision avoidance rules consists of a multi-source data acquisition and preprocessing module, an encounter scenario recognition module, a collision avoidance rule inference engine, a collision risk quantification assessment module, a graded early warning and collision avoidance decision generation module, and a human-computer interaction module, which are connected in sequence.

[0025] The multi-source data acquisition and preprocessing module collects data through the ship's existing sensing equipment: it acquires dynamic and static data such as the ship's name, position, course, speed, and length of the ship and surrounding target ships through the AIS receiver; it acquires distance and bearing echo data of target ships through navigation radar to supplement non-cooperative targets missing from the AIS signal; it acquires static data such as waterways, obstructions, navigation area delineation, and water depth through the electronic chart system; it acquires environmental data such as wind speed and direction, current speed and direction, and visibility through the ship's weather instrument and Doppler log; and it retrieves maneuvering parameters such as turning radius, braking distance, rudder response delay, and main engine acceleration characteristics of the ship under different load and draft conditions from the ship's maneuvering performance database. During preprocessing, linear interpolation and WGS-84 coordinate transformation algorithms were used to unify all data to the same timestamp and geodetic coordinate system to complete spatiotemporal registration. The 3σ criterion was used to remove anomalous jump data, and adaptive Kalman filtering was used to complete and smooth missing and noisy data. For targets with the same source as AIS and radar, a weighted fusion algorithm was used to calculate the target's final position and speed. In open water, AIS data accounted for 70% of the weight and radar data accounted for 30%. In poor visibility scenarios, the weight was automatically adjusted to radar data accounting for 70%. Finally, a standardized navigation dataset was output.

[0026] After receiving the standardized navigation dataset, the encounter scene recognition module calculates the real-time bearing, relative distance, relative velocity vector, and heading angle of the target vessel relative to the current vessel through planar coordinate transformation. Based on the scene definitions in COLREGs, a hierarchical judgment rule is set: visibility conditions are prioritized; if the visibility distance is less than 1 nautical mile, it is directly judged as a poor visibility restricted scene. When visibility is good, overtaking, face-to-face, and cross-encounter situations are judged sequentially. The overtaking situation is judged when the current vessel approaches from a direction greater than 22.5° aft of the target vessel's beam and its speed is greater than the target vessel's. The face-to-face situation is judged when the heading difference between the two vessels is between 165° and 195° and they approach each other with opposite headings. Other scenarios where approaching each other with a collision risk are judged as cross-encounter situations. When multiple target vessels are present, the module performs scene judgment for each target individually and initially sorts them by risk from lowest to highest TCPA (Transient Threshold Allocation).

[0027] The collision avoidance rule reasoning engine's built-in rule knowledge base stores all core clauses of COLREGs using an "IF-THEN" production rule structure. This covers various encounter situations in mutual visibility, rules for actions in poor visibility, and special rules for narrow channels and lane separation. For example: IF: Encountering situation AND Mutual visibility THEN Both vessels have the responsibility to avoid each other, each turning to starboard and passing the other vessel on its port side; IF: Encountering situation AND This vessel is the giving way vessel THEN This vessel should turn to starboard and pass the other vessel's stern, and is prohibited from crossing the other vessel's bow. The rule knowledge base adopts a three-layer architecture, from top to bottom: the basic encounter rule layer, the special navigation area rule layer, and the exception rule layer. The basic encounter rule layer stores general encounter and avoidance rules under mutual visibility and poor visibility conditions; the special navigation area rule layer stores supplementary rules for specific waters such as narrow channels, lane separation areas, and fishing areas; the exception rule layer stores departure rules under imminent danger and special exemptions for vessels with limited maneuverability. During inference execution, a hierarchical matching logic is followed: "special rules take precedence over general rules, visibility clauses take precedence over mutual visibility clauses, and yielding responsibility takes precedence over straight-ahead rights," triggering corresponding clauses layer by layer. When rule conflicts exist, they are resolved according to the above priorities. For complex boundary scenarios, the rule base has built-in special judgment rules, such as: IF During overtaking, the relative bearings of the two vessels gradually enter the cross-ahead section; THEN Overtaking responsibility remains valid, and this vessel still bears full yielding responsibility, without changing to a cross-encounter situation due to changes in relative bearings. This effectively solves the problem of liability determination bias under traditional simple rule matching. The inference engine ultimately outputs the vessel's yielding / straight-ahead responsibility characterization, turning / deceleration requirements, action range constraints, and a list of prohibited behaviors.

[0028] The collision risk quantification assessment module first calculates the basic DCPA and TCPA based on the relative motion vector method. Then, it calculates the ship's trajectory deviation based on wind, current, and pressure data to correct the target's relative motion trajectory. Finally, it combines the ship's turning radius and emergency braking distance parameters to correct the effective encounter distance and available reaction time under emergency avoidance conditions. This module uses the corrected DCPA and TCPA as core indicators, combined with avoidance responsibility weights, target ship size weights, and environmental interference weights, and employs a fuzzy comprehensive evaluation method to calculate the comprehensive collision risk value. Specifically, it uses a triangular membership function to construct the membership matrix of each evaluation indicator, selects a weighted average fuzzy operator to complete the fuzzy calculation, and outputs the comprehensive collision risk value in the 0-1 interval. The risk thresholds are set according to the standard for merchant ships in open waters: a risk value of 0-0.25 corresponds to low risk, with a adjusted DCPA greater than 2 nautical miles and a TCPA greater than 15 minutes; 0.25-0.5 corresponds to medium risk, with a adjusted DCPA of 1-2 nautical miles and a TCPA of 8-15 minutes; 0.5-0.75 corresponds to high risk, with a adjusted DCPA of 0.5-1 nautical miles and a TCPA of 4-8 minutes; and 0.75-1 corresponds to emergency risk, with a adjusted DCPA less than 0.5 nautical miles and a TCPA less than 4 minutes. These thresholds can be adaptively adjusted according to vessel type, load condition, and navigation area.

[0029] The graded early warning and collision avoidance decision generation module triggers graded early warnings based on risk levels: low risk only displays a persistent prompt on the interface; medium risk pops up a text prompt accompanied by a low-frequency audio-visual prompt; high risk pops up a forced pop-up window accompanied by a high-frequency audio-visual prompt; and emergency risk triggers an audio-visual alarm across the entire bridge. Collision avoidance decision generation uses rule constraints as the feasible domain boundary and employs a particle swarm optimization algorithm to solve for the optimal avoidance scheme. When three or more target vessels are present simultaneously, the system first calculates the pairwise collision risk value for each vessel to construct a global risk matrix. A genetic algorithm is then used to solve for a multi-target global avoidance strategy. Under the premise of satisfying the collision avoidance rule constraints for all targets, the system finds the scheme with the lowest overall risk in a single avoidance action, ensuring that no new high-risk encounters with third-party vessels are formed during the avoidance process, thus avoiding the problem of neglecting certain aspects due to traditional single-vessel sequential avoidance. The algorithm sets three optimization objectives: first, reducing the collision risk value after avoidance to below the safety threshold, accounting for 60% of the weight; second, minimizing the heading deviation and range increment caused by the avoidance maneuver, accounting for 25% of the weight; and third, ensuring that the avoidance maneuver's magnitude conforms to the collision avoidance principle of "early, large, wide, and clear," accounting for 15% of the weight. Simultaneously, clear action constraints are set: the turning angle ranges from 15° to 60°, and a single speed adjustment cannot exceed 30% of the current speed, ensuring that the avoidance maneuver complies with the rules and is maneuverable. The final output includes a quantitative scheme containing the turning direction and angle, speed adjustment amount, and optimal execution timing. After generating the plan, the MMG ship motion mathematical model is invoked, and the ship's maneuvering parameters and environmental disturbances are input. The avoidance action is broken down into three stages: rudder command issuance, ship response turning, and entering a new course to maintain course. The trajectory changes are simulated second by second to continuously verify the real-time collision risk on the simulated trajectory, ensuring that the risk remains within a safe range throughout the entire avoidance process. If the conditions are not met, the parameters are adjusted and iteratively optimized. At the same time, it is verified whether the avoidance plan will create new collision hazards with other surrounding targets, and a globally feasible final plan is output.

[0030] The human-machine interface module is integrated into the electronic chart display terminal on the bridge, employing a layered design: the base map is the electronic chart, overlaid with the target vessel's position, heading vector, encounter scenario markings, risk warning areas, and recommended collision avoidance routes; the sidebar displays real-time risk levels, rule liability determination results, and collision avoidance operation suggestions. The pilot can customize warning thresholds and applicable navigation areas through the interface, or manually input avoidance intentions, which are then verified for compliance and safety by the system. The module also supports historical navigation data playback, allowing for post-encounter analysis of past encounters, reconstructing the vessel's situation, risk changes, and system decision-making processes. This can be used for crew collision avoidance training and accident cause tracing. The system reserves a standard data interface for integration with the vessel's autopilot and integrated navigation systems, enabling semi-automatic issuance of avoidance commands after pilot authorization.

[0031] To verify the accuracy of the system's early warning and the compliance of its decision-making, this system conducts simulation tests on typical scenarios using a ship maneuvering simulator, covering four typical navigation scenarios: head-on encounters, cross-encounter encounters, multi-ship encounters, and poor visibility.

[0032] Test results show that, compared with traditional threshold-based collision warning systems, this system reduces the false alarm rate by about 35% and the missed alarm rate by about 28% in complex sea conditions and multi-target scenarios. The generated collision avoidance decision schemes have a compliance rate of over 92% with international maritime collision avoidance rules, and the vessels can maintain a stable safe encounter distance after collision avoidance, effectively verifying the technical effectiveness and practical value of the system.

[0033] The above description is only a preferred embodiment of the present invention and is not intended to limit the scope of the present invention. All equivalent changes and modifications made in accordance with the claims of the present invention should be included within the scope of the present invention.

Claims

1. A ship collision early warning system based on collision avoidance rules, characterized in that, include: The multi-source data acquisition and preprocessing module is used to collect navigation perception data, environmental data and ship handling parameters of the ship and the target ship, perform spatiotemporal registration, outlier data removal and data completion, and output a standardized navigation dataset. The encounter scene recognition module is used to calculate the bearing, distance, relative speed and heading difference of the target ship relative to the ship based on the standardized navigation dataset, and automatically identify the encounter scene type according to the maritime collision avoidance rules; The collision avoidance rule reasoning engine has a built-in collision avoidance rule knowledge base, which is used to map the identified encounter scenario types to match the corresponding collision avoidance rule clauses, and output the division of avoidance responsibility and avoidance action constraints between the ship and the target ship. The collision risk quantification assessment module is used to combine the encounter scenario type and avoidance responsibility, calculate the corrected nearest encounter distance and nearest encounter time, introduce ship handling parameters and environmental disturbances to construct a risk assessment model, and output the collision risk level. The graded early warning and collision avoidance decision generation module is used to trigger early warning signals of the corresponding level according to the collision risk level, generate quantitative collision avoidance action plans based on avoidance action constraints, and deduce the navigation trajectory after the collision avoidance action to verify the safety of the plan. The human-computer interaction module is used to display the encounter situation, collision risk level, early warning information and collision avoidance action plan in real time, and to receive manual intervention instructions.

2. The ship collision early warning system based on collision avoidance rules according to claim 1, characterized in that, The navigation perception data collected by the multi-source data acquisition and preprocessing module includes AIS data, radar detection data, and electronic chart data. Environmental data includes wind speed, wind direction, current speed, current direction, and visibility data. Ship maneuvering parameters include the ship's turning radius, braking distance, and maneuvering response delay parameters. The preprocessing module uses a spatiotemporal interpolation algorithm to achieve unified registration of timestamps and spatial coordinates of the multi-source data.

3. The ship collision early warning system based on collision avoidance rules according to claim 1, characterized in that, The encounter scene recognition module identifies encounter scene types including face-to-face encounters, cross-encounter encounters, overtaking encounters, and poor visibility scenarios. Among them, the conditions for determining a face-to-face encounter are that the difference in heading between the target ship and the ship is in the range of 165°~195°, and the conditions for determining an overtaking encounter are that the target ship is at least 22.5° aft of the ship's main beam and the ship's speed is greater than that of the target ship.

4. The ship collision early warning system based on collision avoidance rules according to claim 1, characterized in that, The collision avoidance rules knowledge base is built on the International Regulations for Preventing Collisions at Sea, 1972. It uses a production rule representation to store the yielding responsibilities, code of conduct, and prohibited behaviors corresponding to each encounter scenario. The collision avoidance rules reasoning engine uses a forward reasoning mechanism to trigger the corresponding rule clauses based on the input encounter scenario and the ship's situation.

5. The ship collision early warning system based on collision avoidance rules according to claim 1, characterized in that, The correction process of the collision risk quantification assessment module is as follows: the relative motion trajectory is corrected based on the offset of the ship's track by the wind pressure, and the calculation results of the encounter distance and encounter time are corrected by combining the ship's turning and braking performance parameters; the risk assessment model adopts the fuzzy comprehensive evaluation method, and outputs a four-level collision risk level with the corrected nearest encounter distance, nearest encounter time and avoidance responsibility weight as input.

6. The ship collision early warning system based on collision avoidance rules according to claim 1, characterized in that, The warning levels of the graded warning and collision avoidance decision generation module include low-risk warning, medium-risk warning, high-risk warning and emergency risk warning, which correspond to warning methods such as sound and light prompts, text prompts, forced pop-ups and sound and light alarm combinations, respectively.

7. The ship collision warning system based on collision avoidance rules according to claim 1 or 6, characterized in that, The quantified collision avoidance action plan includes the turning direction and angle, speed adjustment amount, and optimal timing of action; the trajectory simulation uses a mathematical model of ship motion to simulate the navigation trajectory after the collision avoidance action is executed, and verifies whether the collision risk under the new trajectory has been reduced to below the safety threshold.

8. The ship collision early warning system based on collision avoidance rules according to claim 1, characterized in that, The human-computer interaction module is integrated into the ship's bridge display terminal, which supports the overlay display of the target ship's situation, risk areas and recommended collision avoidance routes on electronic charts, and allows the driver to manually set warning thresholds and rule adaptation parameters.

Images