A partially centralized coordination method for achieving collision avoidance among multiple vessels through mediation

Abstract

This invention provides a partially centralized coordination method for multi-ship collision avoidance through mediation, comprising: calculating ship kinematic parameters; quantifying the collision risk between two ships based on relative motion parameters; comprehensively considering the constraints of COLREGs on the encounter situation; modeling the interaction topology of the multi-ship encounter situation as a dynamic undirected graph; constructing a dynamic topology graph containing interaction and communication relationships; formally modeling the multi-ship collision avoidance problem as a distributed constrained optimization problem; establishing a quintuple model consisting of a set of ship agents, a set of decision variables, a set of discrete value ranges, a set of constraints, and a set of cost functions; introducing an optimal asynchronous partial coverage algorithm; dynamically selecting mediators with the goal of minimizing global cost; performing centralized optimization in local high-risk subgraphs; and obtaining a coordinated collision avoidance scheme. This invention achieves a balance between low communication requirements and high solution quality, effectively improving the cooperative collision avoidance capability between ships, reducing collision risk, and improving maritime traffic safety.

Description

Technical Field

[0001] This invention relates to the field of intelligent navigation and collaborative collision avoidance decision-making technology for ships, and more particularly to a method for partially centralized coordination of multi-ship collision avoidance through mediation. Background Technology

[0002] Ship collisions are a core threat to maritime navigation safety, potentially causing irreversible damage to crew lives, shipping economics, and the marine ecosystem. With the development of intelligent navigation technology, the ability of ships to autonomously avoid collisions has become a crucial element in ensuring navigational safety. Current mainstream collision avoidance algorithms (such as speed barrier methods, artificial potential field methods, or reinforcement learning methods) are typically based on local path planning from a single ship's perspective. This isolated decision-making model is prone to failure in complex water environments with high ship density and strong motion coupling. Frequent maneuvers by individual vessels not only make effective collision avoidance difficult but can also trigger chain reactions, leading to overall traffic flow disruptions and even secondary accidents.

[0003] Existing cooperative collision avoidance technologies primarily employ centralized or distributed architectures. Centralized solutions calculate the globally optimal solution through a central control node, but their effectiveness is limited by the stability of the communication link, inherently suffering from single-point-of-failure risks and insufficient real-time performance. While distributed solutions improve system robustness through local decision-making, they are prone to decision oscillations and low coordination efficiency in dynamic multi-ship interaction scenarios. Of particular note is the significant reduction in inter-ship cooperative efficiency due to the uncertainty of the maritime communication environment. This contradiction between information transmission delays and the real-time requirements of collision avoidance essentially reflects a structural conflict between global optimization goals and individual autonomous decision-making capabilities. Current technologies have not effectively resolved this contradiction, making it difficult to simultaneously achieve both reliability and timeliness in multi-ship cooperative collision avoidance in densely populated waters. Summary of the Invention

[0004] To address the technical problems of existing centralized methods (high communication costs and reliance on central nodes) and distributed methods (low coordination efficiency under complex constraints), this invention provides a partially centralized coordination method for multi-ship collision avoidance through mediation. This invention primarily utilizes a partially centralized coordination mechanism in distributed constraint optimization problems. By dynamically selecting mediators to perform centralized optimization in local high-risk subgraphs, it achieves high-quality coordinated collision avoidance with low communication requirements.

[0005] The technical means employed in this invention are as follows:

[0006] A partially centralized coordination method for achieving collision avoidance among multiple vessels through mediation includes: S1. Calculate the kinematic parameters of the ships, quantify the collision risk between the two ships based on the relative motion parameters, and comprehensively consider the constraints of the International Regulations for Preventing Collisions at Sea on the encounter situation. Model the interactive topology of the multi-ship encounter situation as a dynamic undirected graph and construct a dynamic topology graph that includes interaction and communication relationships. S2. The multi-ship collision avoidance problem is formally modeled as a distributed constrained optimization problem, and a five-tuple model is established, consisting of a set of ship agents, a set of decision variables, a set of discrete value ranges, a set of constraints, and a set of cost functions. S3. Introduce the Optimal Asynchronous Partial Coverage Algorithm (OptAPO), which dynamically selects mediators with the goal of minimizing global cost and performs centralized optimization in local high-risk subgraphs to obtain coordinated collision avoidance schemes.

[0007] Further, step S1 includes: S11. Obtain the kinematic parameters of the ship and other ships, including the position coordinates, heading angle, and speed vector of the ship and other ships, and calculate the relative motion parameters between the two ships, including the relative position vector, relative speed vector, relative heading, nearest encounter distance, and time to reach the nearest encounter point. S12. Based on the relative motion parameters between the two ships, the overall collision risk of the two ships is quantified. The overall collision risk integrates spatial dimension risk and temporal dimension risk. S13. Determine the encounter situation and the liability relationship between the give-way vessel and the vessel proceeding in the straight course in accordance with the International Regulations for Preventing Collisions at Sea; S14. Establish communication link rules based on restricted communication awareness and construct a dynamic undirected topology graph.

[0008] Further, step S13 includes: S131. Determine the type of encounter based on the relative bearing of the two ships, including face-to-face encounter, cross encounter, and overtaking encounter; S132. In a cross-road situation, the vessel located in front of the other vessel on the main beam and whose course is crossed is determined to be the vessel giving way. S133. In an overtaking situation, the overtaking vessel that is located directly behind another vessel and has the same course is determined to be the yielding vessel. S134. In a head-on situation, both vessels are yielding vessels and each turns to starboard to give way.

[0009] Further, step S14 includes: S141. Abstract each ship as a graph node and establish a ship set; S142. Set the detection radius as the threshold for establishing a communication link. When the distance between the two ships is less than the detection radius, an undirected edge is established to form an edge set. S143. Each ship only perceives the state information of its neighboring nodes and exchanges rudder angle and cost information with its neighboring nodes. S144. The node positions and edge set connections are updated in real time with the movement of the ship, realizing the dynamic evolution of the topology.

[0010] Further, step S2 includes: S21. Based on the constructed dynamic topology graph, each ship node is abstracted as an intelligent agent, and a set of ship intelligent agents is established. Each intelligent agent maintains a local state including the current position, heading, speed, and rudder angle, and has the independent decision-making ability to calculate costs and perform collision avoidance actions based on the local view. Intelligent agents establish constraint relationships through edge sets and exchange information only with neighboring intelligent agents. S22. Assign a rudder angle as a decision variable to each agent and establish a set of decision variables; the value of the decision variable determines the change in the ship's course at the next moment. The decision variables of each agent are independent of each other and are related by a set of constraints. S23. Determine the physical limits of the maximum left rudder angle and the maximum right rudder angle, set the discrete step length to uniformly discretize the continuous rudder angle space into a finite number of discrete rudder angle values, and establish a discrete value range set that provides a finite number of possible values ​​for each decision variable. S24. Establish a set of constraints, including a distance constraint that requires the distance between any two ships to be greater than a safe distance threshold and a maneuvering constraint that requires the change in the ship's course to be less than the maximum course change limit. The safe distance threshold is determined based on the ship's dimensions and the navigation environment, and the maximum course change limit is determined based on the ship's maneuverability. Both the distance constraint and the maneuvering constraint are hard constraints. S25. Establish a set of cost functions, including local cost functions and global costs. The local cost functions are calculated independently by each agent and are used to quantify the safety, rule compliance and navigation efficiency costs of the agent under a specific rudder angle. The global cost is the weighted sum of the local costs of all agents and is used to evaluate the optimization objective of the entire system, with minimization as the optimization objective.

[0011] Further, in step S25, the local cost function is constructed as a weighted sum, comprehensively evaluating security costs, COLREGs compliance costs, and economic efficiency costs based on satisfying the hard constraint set, wherein: The safety cost incorporates risks from both spatial and temporal dimensions. An exponential function is used to construct a high-gradient repulsive field to ensure sufficient safety margin. The specific calculation method is as follows:

[0012] in, Indicates the first The safety cost of a ship Indicates the first Decision variables for a ship Indicates the first The set of decision variables of the neighboring ships of a vessel. Indicates the first The neighbors of the ship gathered. Indicates traversing the first... All neighboring vessels of the vessel This represents the safety weighting coefficient. Represents an exponential function. Indicates the risk sensitivity coefficient. Indicates the first The ship and the first The risk of collision between neighboring vessels Indicates an indicator function, Indicates the risk perception threshold; COLREGs compliance costs are quantitative penalties for soft constraints, assessed based on the give-way vessel's responsibility and whether the maneuvering was port-handle. The specific calculation method is as follows:

[0013] in, Indicates the first COLREGs compliance costs for a single vessel This represents the rule compliance weighting coefficient. Indicates the conditions for determining which vessel should give way. This indicates the conditions for determining left-hand drive control. Indicates the first the rudder angle of a ship, This represents the logical AND operator; Economic efficiency cost is used to penalize deviations from the predicted course and the target course, as well as large rudder angle operations, to ensure that the vessel maintains its original course as much as possible while ensuring safety and compliance. The specific calculation method is as follows:

[0014] in, Indicates the first The economic efficiency cost of a ship This represents the heading-maintaining weighting coefficient, used to penalize deviations between the predicted and target headings. Indicates the predicted course after steering. Indicates the ship's target course. This indicates the change in heading.

[0015] Further, step S3 includes: S31. Each ship initializes its local state, including initial rudder angle, local optimal cost, and mediation intention; it adds itself to the list of constraints it maintains and broadcasts these states to its neighbors via initialization messages; based on the magnitude of the collision risk, the priority of ships is dynamically adjusted during the algorithm's operation, assigning the mediator status to ships with higher collision risk; when a ship continuously and asynchronously receives state query messages or confirmation messages from its neighbors, it updates the state in its local view. S32. The ship calculates the current cost of the local subgraph. If the current cost is greater than the local optimal cost, a conflict is determined. The ship intending to mediate will first try to eliminate the conflict by changing its own rudder angle. If it fails and does not receive an active intention from a higher priority ship, it will initiate an active conversation or directly initiate passive mediation. S33. The mediator sends an assessment request to the ships in the constraint relationship list to collect feedback costs. The centralized search is performed using the branch and bound method, with the mediator's action set as the root node and other ship actions as branches. Branches that do not meet the constraint conditions are pruned to minimize the internal ship conflict cost. S34. If the new solution generated by the active mediation does not worsen the global state, the mediator sends a confirmation message to the participating ships to agree to their adoption of the new rudder angle; if the new solution leads to an increase in the cost of external ships, an initialization message is sent to include them in the view in future iterations; if the new solution worsens the global state, the mediator abandons the local optimum and retains the current state.

[0016] Further, step S32 includes: S321. The ship calculates the current cost of the local subgraph using the current rudder angle information of the known neighbors. If the current cost is greater than the local optimal cost, it indicates that there is a conflict. S322. A vessel intending to initiate mediation shall first attempt to resolve the conflict by simply changing its own rudder angle. If it is unable to achieve the goal without affecting its lower-priority neighbors and has not received any initiative from a higher-priority vessel, it shall initiate an active conversation as a mediator. S323. Passive mediation is initiated directly by the vessel to respect the decision-making logic of high-priority vessels and update local expectations without forcibly changing the current system solution.

[0017] Further, step S33 includes: S331. The vessel designated as the mediator sends an assessment request to the relevant vessels in its list of constraints and collects feedback costs for each candidate rudder angle. S332. The mediator uses the branch and bound method to perform a centralized search, with the mediator's action set as the root node, and the branches extending out are the actions of other ships. Branches that do not meet the safety distance constraint and maneuvering limit constraint are pruned, and the combination of rudder angles that minimizes the local conflict cost is searched in the remaining feasible solution space.

[0018] Compared with the prior art, the present invention has the following advantages: 1. The partially centralized coordination method for multi-ship collision avoidance provided by this invention combines with the partially centralized coordination mechanism for distributed constraint optimization problems. It adopts a coordination architecture of local centralization and global distribution, and achieves high-quality collision avoidance decision-making under low communication requirements. It effectively solves the technical problems of high communication cost and dependence on central nodes in centralized methods and low coordination efficiency in distributed methods under complex constraints.

[0019] 2. This invention achieves the goal of minimizing global cost by dynamically selecting mediators and performing centralized optimization in local high-risk subgraphs. Cost tables are exchanged only between mediators and key neighbors. Compared with fully distributed solution algorithms, this reduces the amount of communication by several orders of magnitude, greatly avoiding communication congestion and delay in restricted maritime networks.

[0020] 3. This invention introduces a weighted evaluation mechanism of safety cost, COLREGs compliance cost and economic efficiency cost, and prioritizes the avoidance method that complies with the International Maritime Collision Prevention Regulations while ensuring the safety of ship collision avoidance, and maintains the original course as much as possible, thus achieving multi-objective optimization of safety, compliance and economy.

[0021] 4. In highly coupled and complex encounter scenarios, this invention can guide ships to make orderly fine adjustments to the global perspective before approaching high-risk areas, reducing abrupt changes in rudder angle. Even in extremely congested conditions, localized optimization can prioritize the handling of the most threatening targets, ensuring that the global minimum distance is always above the safety threshold, achieving smooth and coordinated collision-free avoidance.

[0022] In summary, the technical solution of this invention, by combining with a partially centralized coordination mechanism for distributed constrained optimization problems, adopts a dynamic adjustment strategy to replace the traditional fully centralized or fully distributed decision-making model. This solves the technical problems of high communication costs and reliance on a central node in centralized methods, and low coordination efficiency in distributed methods under complex constraints in existing technologies. Therefore, the technical solution of this invention solves the technical problems of high communication overhead, low coordination efficiency, and high risk of single point of failure in existing technologies.

[0023] Based on the above reasons, this invention can be widely applied in fields such as intelligent ship navigation, maritime traffic management systems, autonomous ship collaborative control, and port ship scheduling. Attached Figure Description

[0024] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0025] Figure 1 This is an overall flowchart of a partially centralized coordination method for achieving collision avoidance among multiple vessels through mediation, according to the present invention.

[0026] Figure 2 This is a schematic diagram of the multi-ship dynamic interactive topology structure of the present invention.

[0027] Figure 3 This is a schematic diagram of the solution process for distributed constraint optimization in this invention.

[0028] Figure 4 This is a timing diagram of asynchronous communication and message processing based on the OptAPO algorithm of this invention.

[0029] Figure 5 This is a tree diagram illustrating the branching and delimiting principle of the mediator's centralized search in this invention. Detailed Implementation

[0030] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0031] It should be noted that the terms "comprising" and "having" and any variations thereof in the specification, claims and accompanying drawings of this invention are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such process, method, product or device.

[0032] like Figure 1 As shown, the present invention provides a partially centralized coordination method for achieving collision avoidance among multiple vessels through mediation, comprising: S1. Calculate the kinematic parameters of the ships, quantify the collision risk between the two ships based on the relative motion parameters, comprehensively consider the constraints of the International Regulations for Preventing Collisions at Sea (COLREGs) on the encounter situation, model the interactive topology of the multi-ship encounter situation as a dynamic undirected graph, and construct a dynamic topology graph containing interaction and communication relationships. S2. The multi-ship collision avoidance problem is formally modeled as a distributed constrained optimization problem (DCOP), and a five-tuple model is established, consisting of a set of ship agents, a set of decision variables, a set of discrete value ranges, a set of constraints, and a set of cost functions. S3. Introduce the Optimal Asynchronous Partial Coverage Algorithm (OptAPO), which dynamically selects mediators with the goal of minimizing global cost and performs centralized optimization in local high-risk subgraphs to obtain coordinated collision avoidance schemes.

[0033] In a specific implementation, as a preferred embodiment of the present invention, step S1 includes: S11. Obtain the kinematic parameters of the ship and other ships, including the position coordinates, heading angle, and speed vector of the ship and other ships, and calculate the relative motion parameters between the two ships, including the relative position vector, relative speed vector, relative heading, nearest encounter distance, and time to reach the nearest encounter point. S12. Based on the relative motion parameters between the two ships, the overall collision risk of the two ships is quantified. The overall collision risk integrates spatial dimension risk and temporal dimension risk. S13. Determine the encounter situation and the liability relationship between the give-way vessel and the vessel proceeding in the straight course in accordance with the International Regulations for Preventing Collisions at Sea; S14. Establish communication link rules based on restricted communication awareness and construct a dynamic undirected topology graph.

[0034] In a specific implementation, as a preferred embodiment of the present invention, step S13 includes: S131. Determine the type of encounter based on the relative bearing of the two ships, including face-to-face encounter, cross encounter, and overtaking encounter; S132. In a cross-road situation, the vessel located in front of the other vessel on the main beam and whose course is crossed is determined to be the vessel giving way. S133. In an overtaking situation, the overtaking vessel that is located directly behind another vessel and has the same course is determined to be the yielding vessel. S134. In a head-on situation, both vessels are yielding vessels and each turns to starboard to give way.

[0035] In a specific implementation, as a preferred embodiment of the present invention, step S14 includes: S141, such as Figure 2 The diagram shows a dynamic interactive topology structure of multiple ships, where each ship is abstracted as a graph node to establish a ship set. S142. Set the detection radius as the threshold for establishing a communication link. When the distance between the two ships is less than the detection radius, an undirected edge is established to form an edge set. S143. Each ship only perceives the state information of its neighboring nodes and exchanges rudder angle and cost information with its neighboring nodes. S144. The node positions and edge set connections are updated in real time with the movement of the ship, realizing the dynamic evolution of the topology.

[0036] In a specific implementation, as a preferred embodiment of the present invention, step S2 includes: S21. Based on the constructed dynamic topology graph, each ship node is abstracted as an intelligent agent, and a set of ship intelligent agents is established. Each intelligent agent maintains a local state including the current position, heading, speed, and rudder angle, and has the independent decision-making ability to calculate costs and perform collision avoidance actions based on the local view. Intelligent agents establish constraint relationships through edge sets and exchange information only with neighboring intelligent agents. S22. Assign a rudder angle as a decision variable to each agent and establish a set of decision variables; the value of the decision variable determines the change in the ship's course at the next moment. The decision variables of each agent are independent of each other and are related by a set of constraints. S23. Determine the physical limits of the maximum left rudder angle and the maximum right rudder angle, set the discrete step length to uniformly discretize the continuous rudder angle space into a finite number of discrete rudder angle values, and establish a discrete value range set that provides a finite number of possible values ​​for each decision variable. S24. Establish a set of constraints, including a distance constraint that requires the distance between any two ships to be greater than a safe distance threshold and a maneuvering constraint that requires the change in the ship's course to be less than the maximum course change limit. The safe distance threshold is determined based on the ship's dimensions and the navigation environment, and the maximum course change limit is determined based on the ship's maneuverability. Both the distance constraint and the maneuvering constraint are hard constraints. S25. Establish a set of cost functions, including local cost functions and global costs. The local cost functions are calculated independently by each agent and are used to quantify the safety, rule compliance and navigation efficiency costs of the agent under a specific rudder angle. The global cost is the weighted sum of the local costs of all agents and is used to evaluate the optimization objective of the entire system, with minimization as the optimization objective.

[0037] In a specific implementation, as a preferred embodiment of the present invention, in step S25, the local cost function is constructed as a weighted sum, comprehensively evaluating security cost, COLREGs compliance cost, and economic efficiency cost based on satisfying the hard constraint set, wherein: The safety cost incorporates risks from both spatial and temporal dimensions. An exponential function is used to construct a high-gradient repulsive field to ensure sufficient safety margin. The specific calculation method is as follows:

[0038] in, Indicates the first The safety cost of a ship Indicates the first Decision variables for a ship Indicates the first The set of decision variables of the neighboring ships of a vessel. Indicates the first The neighbors of the ship gathered. Indicates traversing the first... All neighboring vessels of the vessel This represents the security weighting coefficient, which has the highest magnitude and ensures that compliance and economy are not sacrificed for security under any circumstances. Represents an exponential function. It represents the risk sensitivity coefficient, which can control the steepness of the risk "potential field". Indicates the first The ship and the first The risk of collision between neighboring vessels Indicates an indicator function, This represents the risk perception threshold; this function is only activated when the risk to another ship reaches a certain level. COLREGs compliance costs are quantitative penalties for soft constraints, assessed based on the give-way vessel's responsibility and whether the maneuvering was port-handle. The specific calculation method is as follows:

[0039] in, Indicates the first COLREGs compliance costs for a single vessel This represents the rule compliance weight coefficient, which has the second highest magnitude, ensuring that, under safe conditions, avoidance methods that comply with COLREGs are given priority. Indicates the conditions for determining which vessel should give way. This indicates the conditions for determining left-hand drive control. Indicates the first the rudder angle of a ship, This represents the logical AND operator; Economic efficiency cost is used to penalize deviations from the predicted course and the target course, as well as large rudder angle operations, to ensure that the vessel maintains its original course as much as possible while ensuring safety and compliance. The specific calculation method is as follows:

[0040] in, Indicates the first The economic efficiency cost of a ship This represents the heading-maintaining weighting coefficient, used to penalize deviations between the predicted and target headings. Indicates the predicted course after steering. Indicates the ship's target heading (original heading). This indicates the change in heading.

[0041] In specific implementation, as a preferred embodiment of the present invention, such as Figure 3 The diagram illustrates the solution process for distributed constraint optimization. The solution process includes preprocessing (constructing the graph structure and search space), initialization (determining the solution strategy and communication mode), and the algorithm solution stage (performing message propagation and coordination), i.e., step S3, which includes: S31. During the initialization and asynchronous message processing phase, such as Figure 4 As shown, each ship acts as an independent intelligent agent. Each ship initializes its local state, including initial rudder angle, local optimal cost, and mediation intention; it adds itself to the constraint relationship list (good-list) maintained by itself and broadcasts these states to its neighbors through an initialization ("init") message; based on the collision risk level (CRI), the priority of ships is dynamically adjusted during the algorithm's operation, assigning the mediator status to ships with higher collision risk levels; when a ship continuously and asynchronously receives a state query ("value?") message or an acceptance ("accept!") message from its neighbors, it updates the state in its local agent-view. S32. During the conflict detection and mediation triggering phase, the vessel assesses the current cost by checking the local view. If the current cost is greater than the local optimal cost, it indicates that the system is not in an optimal solution state. Vessels intending to actively mediate first attempt to eliminate the conflict by simply changing their own rudder angle. If this fails and no active intention is received from a higher-priority vessel, an active session will be formally initiated. Passive mediation is initiated directly, with the aim of respecting the decision-making logic of higher-priority vessels and updating local expectations, without forcibly changing the current system solution. S33. In the centralized solution and decision distribution phase, the mediator sends evaluation requests (“evaluate?”) to ships in the constraint list (good-list) and collects feedback costs (“evaluate!”) for each candidate rudder angle, such as... Figure 5 As shown, the mediator uses the branch and bound method to perform a centralized search, with the mediator's action set as the root node and other ship actions as branches. Branches that do not meet the constraints are pruned to minimize the internal ship conflict cost. S34. If the new solution generated by the active mediation does not worsen the global state, the mediator sends an acceptance ("accept!") message to the participating ships to agree to their adoption of the new rudder angle; if the new solution leads to an increase in costs for external ships, an initialization ("init") message is sent to incorporate it into the view in future iterations; if the new solution worsens the global state, the mediator abandons the local optimum and retains the current state. Finally, each ship asynchronously broadcasts the final decision to achieve global cooperative collision avoidance.

[0042] In a specific implementation, as a preferred embodiment of the present invention, step S32 includes: S321. The ship calculates the current cost of the local subgraph using the current rudder angle information of the known neighbors. If the current cost is greater than the local optimal cost, it indicates that there is a conflict. S322. A vessel intending to initiate mediation shall first attempt to resolve the conflict by simply changing its own rudder angle. If it is unable to achieve the goal without affecting its lower-priority neighbors and has not received any initiative from a higher-priority vessel, it shall initiate an active conversation as a mediator. S323. Passive mediation is initiated directly by the vessel to respect the decision-making logic of high-priority vessels and update local expectations without forcibly changing the current system solution.

[0043] In a specific implementation, as a preferred embodiment of the present invention, step S33 includes: S331. The vessel acting as mediator sends an assessment request to the relevant vessels in its good-list of obligations and collects feedback costs for each candidate rudder angle. S332. The mediator performs a centralized search using the branch and bound method, with the mediator's action set as the root node and the branches extending from it representing the actions of other ships. Branches that do not meet the safety distance constraint and maneuvering limit constraint are pruned. The mediator searches for the combination of rudder angles that minimizes the cost of local conflict in the remaining feasible solution space.

[0044] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for partial centralized coordination of multi-ship collision avoidance by mediation, characterized in that, include: S1. Calculate the kinematic parameters of the ships, quantify the collision risk between the two ships based on the relative motion parameters, and comprehensively consider the constraints of the International Regulations for Preventing Collisions at Sea on the encounter situation. Model the interactive topology of the multi-ship encounter situation as a dynamic undirected graph and construct a dynamic topology graph that includes interaction and communication relationships. S2. The multi-ship collision avoidance problem is formally modeled as a distributed constrained optimization problem, and a five-tuple model is established, consisting of a set of ship agents, a set of decision variables, a set of discrete value ranges, a set of constraints, and a set of cost functions. S3. Introduce the optimal asynchronous partial coverage algorithm, which dynamically selects mediators with the goal of minimizing global cost, performs centralized optimization in local high-risk subgraphs, and obtains coordinated collision avoidance schemes.

2. A method for partial centralized coordination of multi-ship collision avoidance by mediation according to claim 1, characterized in that, Step S1 includes: S11. Obtain the kinematic parameters of the ship and other ships, including the position coordinates, heading angle, and speed vector of the ship and other ships, and calculate the relative motion parameters between the two ships, including the relative position vector, relative speed vector, relative heading, nearest encounter distance, and time to reach the nearest encounter point. S12. Based on the relative motion parameters between the two ships, the overall collision risk of the two ships is quantified. The overall collision risk integrates spatial dimension risk and temporal dimension risk. S13. Determine the encounter situation and the liability relationship between the give-way vessel and the vessel proceeding in the straight course in accordance with the International Regulations for Preventing Collisions at Sea; S14. Establish communication link rules based on restricted communication awareness and construct a dynamic undirected topology graph.

3. The method for partial centralized coordination of multi-ship collision avoidance by mediation according to claim 2, characterized in that, Step S13 includes: S131. Determine the type of encounter based on the relative bearing of the two ships, including face-to-face encounter, cross encounter, and overtaking encounter; S132. In a cross-road situation, the vessel located in front of the other vessel on the main beam and whose course is crossed is determined to be the vessel giving way. S133. In an overtaking situation, the overtaking vessel that is located directly behind another vessel and has the same course is determined to be the yielding vessel. S134. In a head-on situation, both vessels are yielding vessels and each turns to starboard to give way.

4. The method of claim 2, wherein, Step S14 includes: S141. Abstract each ship as a graph node and establish a ship set; S142. Set the detection radius as the threshold for establishing a communication link. When the distance between the two ships is less than the detection radius, an undirected edge is established to form an edge set. S143. Each ship only perceives the state information of its neighboring nodes and exchanges rudder angle and cost information with its neighboring nodes. S144. The node positions and edge set connections are updated in real time with the movement of the ship, realizing the dynamic evolution of the topology.

5. The method for partial centralized coordination of multi-ship collision avoidance by mediation according to claim 1, wherein, Step S2 includes: S21. Based on the constructed dynamic topology graph, each ship node is abstracted as an intelligent agent, and a set of ship intelligent agents is established. Each intelligent agent maintains a local state including the current position, heading, speed, and rudder angle, and has the independent decision-making ability to calculate costs and perform collision avoidance actions based on the local view. Intelligent agents establish constraint relationships through edge sets and exchange information only with neighboring intelligent agents. S22. Assign a rudder angle as a decision variable to each agent and establish a set of decision variables; the value of the decision variable determines the change in the ship's course at the next moment. The decision variables of each agent are independent of each other and are related by a set of constraints. S23. Determine the physical limits of the maximum left rudder angle and the maximum right rudder angle, set the discrete step length to uniformly discretize the continuous rudder angle space into a finite number of discrete rudder angle values, and establish a discrete value range set that provides a finite number of possible values ​​for each decision variable. S24. Establish a set of constraints, including a distance constraint that requires the distance between any two ships to be greater than a safe distance threshold and a maneuvering constraint that requires the change in the ship's course to be less than the maximum course change limit. The safe distance threshold is determined based on the ship's dimensions and the navigation environment, and the maximum course change limit is determined based on the ship's maneuverability. Both the distance constraint and the maneuvering constraint are hard constraints. S25. Establish a set of cost functions, including local cost functions and global costs. The local cost functions are calculated independently by each agent and are used to quantify the safety, rule compliance and navigation efficiency costs of the agent under a specific rudder angle. The global cost is the weighted sum of the local costs of all agents and is used to evaluate the optimization objective of the entire system, with minimization as the optimization objective.

6. A partially centralized coordination method for achieving collision avoidance among multiple vessels through mediation, as described in claim 5, is characterized in that... In step S25, the local cost function is constructed as a weighted sum, comprehensively evaluating security costs, COLREGs compliance costs, and economic efficiency costs while satisfying the hard constraint set, wherein: The safety cost incorporates risks from both spatial and temporal dimensions. An exponential function is used to construct a high-gradient repulsive field to ensure sufficient safety margin. The specific calculation method is as follows: in, Indicates the first The safety cost of a ship Indicates the first Decision variables for a ship Indicates the first The set of decision variables of the neighboring ships of a vessel. Indicates the first The neighbors of the ship gathered. Indicates traversing the first... All neighboring vessels of the vessel This represents the safety weighting coefficient. Represents an exponential function. Indicates the risk sensitivity coefficient. Indicates the first The ship and the first The risk of collision between neighboring vessels Indicates an indicator function, Indicates the risk perception threshold; COLREGs compliance costs are quantitative penalties for soft constraints, assessed based on the give-way vessel's responsibility and whether the maneuvering was port-handle. The specific calculation method is as follows: in, Indicates the first COLREGs compliance costs for a single vessel This represents the rule compliance weighting coefficient. Indicates the conditions for determining which vessel should give way. This indicates the conditions for determining left-hand drive control. Indicates the first the rudder angle of a ship, This represents the logical AND operator; Economic efficiency cost is used to penalize deviations from the predicted course and the target course, as well as large rudder angle operations, to ensure that the vessel maintains its original course as much as possible while ensuring safety and compliance. The specific calculation method is as follows: in, Indicates the first The economic efficiency cost of a ship This represents the heading-maintaining weighting coefficient, used to penalize deviations between the predicted and target headings. Indicates the predicted course after steering. Indicates the ship's target course. This indicates the change in heading.

7. A partially centralized coordination method for achieving collision avoidance among multiple vessels through mediation, as described in claim 1, is characterized in that... Step S3 includes: S31. Each ship initializes its local state, including initial rudder angle, local optimal cost, and mediation intention; it adds itself to the list of constraints it maintains and broadcasts these states to its neighbors via initialization messages; based on the magnitude of the collision risk, the priority of ships is dynamically adjusted during the algorithm's operation, assigning the mediator status to ships with higher collision risk; when a ship continuously and asynchronously receives state query messages or confirmation messages from its neighbors, it updates the state in its local view. S32. The ship calculates the current cost of the local subgraph. If the current cost is greater than the local optimal cost, a conflict is determined. The ship intending to mediate will first try to eliminate the conflict by changing its own rudder angle. If it fails and does not receive an active intention from a higher priority ship, it will initiate an active conversation or directly initiate passive mediation. S33. The mediator sends an assessment request to the ships in the constraint relationship list to collect feedback costs. The centralized search is performed using the branch and bound method, with the mediator's action set as the root node and other ship actions as branches. Branches that do not meet the constraint conditions are pruned to minimize the internal ship conflict cost. S34. If the new solution generated by the active mediation does not worsen the global state, the mediator sends a confirmation message to the participating ships to agree to their adoption of the new rudder angle; if the new solution leads to an increase in the cost of external ships, an initialization message is sent to include them in the view in future iterations; if the new solution worsens the global state, the mediator abandons the local optimum and retains the current state.

8. A partially centralized coordination method for achieving collision avoidance among multiple vessels through mediation, as described in claim 7, is characterized in that... Step S32 includes: S321. The ship calculates the current cost of the local subgraph using the current rudder angle information of the known neighbors. If the current cost is greater than the local optimal cost, it indicates that there is a conflict. S322. A vessel intending to initiate mediation shall first attempt to resolve the conflict by simply changing its own rudder angle. If it is unable to achieve the goal without affecting its lower-priority neighbors and has not received any initiative from a higher-priority vessel, it shall initiate an active conversation as a mediator. S323. Passive mediation is initiated directly by the vessel to respect the decision-making logic of high-priority vessels and update local expectations without forcibly changing the current system solution.

9. A partially centralized coordination method for achieving collision avoidance among multiple vessels through mediation, as described in claim 7, is characterized in that... Step S33 includes: S331. The vessel designated as the mediator sends an assessment request to the relevant vessels in its list of constraints and collects feedback costs for each candidate rudder angle. S332. The mediator uses the branch and bound method to perform a centralized search, with the mediator's action set as the root node, and the branches extending out are the actions of other ships. Branches that do not meet the safety distance constraint and maneuvering limit constraint are pruned, and the combination of rudder angles that minimizes the local conflict cost is searched in the remaining feasible solution space.

Images