A ship intelligent collision avoidance method and device, electronic equipment and medium

Abstract

The application provides a ship intelligent collision avoidance method and device, electronic equipment and medium, the method comprises: configuring a multi-stage behavior state machine of a target ship; when it is judged that the target ship and other ships in a preset perception range exist a collision risk, the multi-stage behavior state machine is switched to a collision avoidance state, and a two-dimensional parameter space corresponding to the target ship is constructed; the two-dimensional parameter space comprises a plurality of candidate decision schemes, for each candidate decision scheme, the position sequence of the target ship and other ships in a future preset time period is predicted; the collision risk value of the target ship and other ships is calculated based on the position sequence point by point in time, and a target decision scheme is screened out from the candidate decision scheme, so as to provide a decision-transparent obstacle avoidance decision in real time.

Description

Technical Field

[0001] This application relates to the field of information processing technology, and more specifically, to a method, device, electronic device, and medium for intelligent collision avoidance of ships. Background Technology

[0002] The current mainstream ship collision avoidance technologies mainly suffer from the following problems: the decision-making process of most AI-based collision avoidance systems (such as deep learning models) is not transparent, making it difficult for crew members to understand and trust the actions recommended by the system; traditional systems usually only provide a single "optimal" solution, lacking alternative solutions for comparison, which limits the decision-making flexibility of operators; and collision avoidance systems with fixed parameters are difficult to adapt to changes in different ship characteristics and navigation environments, and lack configurability.

[0003] Moreover, complex optimization algorithms involve huge computational demands in complex scenarios involving multiple ships meeting, making it difficult to meet the timeliness requirements of real-time collision avoidance decision-making; especially in high-density traffic waters, computational delays may lead to missing the best collision avoidance opportunity. Summary of the Invention

[0004] In view of this, the purpose of this application is to provide a ship intelligent collision avoidance method, device, electronic device and medium that can provide real-time and transparent obstacle avoidance decisions.

[0005] This application provides an intelligent collision avoidance method for ships, the method comprising: Configure a multi-stage behavioral state machine for the target ship; the multi-stage behavioral state machine includes a normal navigation state, a collision avoidance state, a range recovery state, and a cautious navigation state; different states of the state machine correspond to different navigation strategies. When it is determined that there is a collision risk between the target ship and other ships within the preset perception range, the multi-stage behavioral state machine switches to the collision avoidance state and constructs a two-dimensional parameter space corresponding to the target ship. The two-dimensional parameter space includes a variety of candidate decision schemes, and each candidate decision scheme is a combination of acceleration parameters and turning rate parameters. For each candidate decision scheme in the two-dimensional parameter space, predict the position sequence of the target ship and other ships in the future within a preset time period; For each candidate decision scheme, the collision risk value of the target ship and other ships is calculated point by point in time based on the position sequence; Based on the collision risk value corresponding to each candidate decision scheme, the target decision scheme is selected from the candidate decision schemes.

[0006] In some embodiments, the intelligent collision avoidance method for ships, the step of constructing the two-dimensional parameter space corresponding to the target ship includes: The acceleration parameters are discretized and sampled with a preset acceleration step size to obtain multiple acceleration sample values; The head-turn rate parameter is discretized and sampled with a preset head-turn rate step size to obtain multiple head-turn rate sample values; The acceleration sample value and the head rotation rate sample value are combined in pairs to generate a candidate scheme set containing all combinations.

[0007] In some embodiments, the intelligent collision avoidance method for ships, calculating the collision risk value of the target ship and other ships point-in-time based on the position sequence, includes: For each time point in the position sequence, calculate the distance between this ship and other ships; If the distance is less than a preset distance threshold, then the time point is marked as a dangerous situation between the target ship and the other ship; The collision risk value for each candidate decision scheme is obtained by summing the number of dangerous situations the target ship experiences at all time points and among all other ships.

[0008] In some embodiments, the intelligent collision avoidance method for ships, the step of selecting a target decision scheme from the candidate decision schemes based on the collision risk value corresponding to each candidate decision scheme, includes: Candidate decision schemes whose collision risk values ​​meet preset risk conditions are identified as target decision schemes; When multiple target decision options have the same collision risk value, the priority of the target decision options is determined based on the steering magnitude.

[0009] In some embodiments, the intelligent collision avoidance method for ships further includes: When the trajectory prediction results of the target ship and other ships within the preset perception range are safe within a preset time period in the future, the multi-stage behavior state machine is controlled to be in normal navigation state. When the state machine is in normal navigation state, it determines the normal navigation instruction based on the current heading, current speed and target planned route. In some embodiments, the intelligent collision avoidance method for ships further includes: The two-dimensional parameter space is updated in real time, as are the collision risk values ​​of the target ship and other ships. When the collision risk value between the target ship and other ships meets the preset risk elimination conditions, the state machine is controlled to switch to the range recovery state.

[0010] When the state machine is in the range recovery state, a return command is generated based on the projected azimuth angle from the current position of the target ship to the target planned route.

[0011] In some embodiments, the intelligent collision avoidance method for ships further includes: Record each candidate decision scheme in the candidate scheme set and the corresponding collision risk value, as well as the feedback information during the execution of the target decision scheme.

[0012] In some embodiments, a ship intelligent collision avoidance device is also provided, the device comprising: The configuration module is used to configure the multi-stage behavior state machine of the target ship; the multi-stage behavior state machine includes normal navigation state, collision avoidance state, range recovery state, and cautious navigation state; different states of the state machine correspond to different navigation strategies; The construction module is used to construct a two-dimensional parameter space corresponding to the target ship when it is determined that there is a collision risk between the target ship and other ships within the preset perception range. The two-dimensional parameter space includes multiple candidate decision schemes, and each candidate decision scheme is a combination of acceleration parameters and turning rate parameters. The prediction module is used to predict the position sequence of the target ship and other ships in a future preset time period for each candidate decision scheme in the two-dimensional parameter space. The calculation module is used to calculate the collision risk value of the target ship and other ships on a time-by-time basis based on the position sequence for each candidate decision scheme; The filtering module is used to filter out the target decision scheme from the candidate decision schemes based on the collision risk value corresponding to each candidate decision scheme.

[0013] In some embodiments, an electronic device is also provided, comprising: a processor, a memory, and a bus, wherein the memory stores machine-readable instructions executable by the processor, and when the electronic device is running, the processor communicates with the memory via the bus, and when the machine-readable instructions are executed by the processor, the steps of the intelligent collision avoidance method for ships are performed.

[0014] In some embodiments, a computer-readable storage medium is also provided, on which a computer program is stored, which, when executed by a processor, performs the steps of the intelligent collision avoidance method for ships.

[0015] This application provides a ship intelligent collision avoidance method, which configures a multi-stage behavioral state machine for the target ship. The multi-stage behavioral state machine includes a normal navigation state, a collision avoidance state, a range recovery state, and a cautious navigation state. Different states of the state machine correspond to different navigation strategies. When it is determined that there is a collision risk between the target ship and other ships within a preset perception range, the multi-stage behavioral state machine switches to the collision avoidance state, constructing a two-dimensional parameter space corresponding to the target ship. The two-dimensional parameter space includes multiple candidate decision schemes, each of which is a combination of acceleration parameters and turning rate parameters. For each candidate decision scheme in the two-dimensional parameter space, the position sequence of the target ship and other ships within a preset time period is predicted. The other ships are vessels within the preset perception range of the target ship. For each candidate decision scheme, based on the... The method calculates the collision risk value of the target ship and other ships at each time point of the position sequence; based on the collision risk value corresponding to each candidate decision scheme, the target decision scheme is selected from the candidate decision schemes; different states of the state machine of the method correspond to different navigation strategies, and the full search of the two-dimensional parameter space is only initiated in the collision avoidance state. In the normal navigation state, no candidate scheme calculation is required, which avoids the waste of computing resources caused by a single decision process running in the entire time domain. This enables the system to run in the most suitable computing mode in different navigation scenarios, which is particularly suitable for shipborne embedded environments with limited computing resources, while reducing system power consumption; the full search ensures that the global optimal solution can be found, and the collision risk value is calculated at each time point to capture the entire process of risk evolution over time. It is suitable for high-density traffic waters, the decision-making process is transparent, and a wide variety of solutions are provided. Attached Figure Description

[0016] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 A flowchart of the intelligent collision avoidance method for ships according to an embodiment of this application is shown; Figure 2 A flowchart illustrating the method for constructing the two-dimensional parameter space corresponding to the target ship according to an embodiment of this application is shown. Figure 3 A flowchart illustrating the method for calculating the collision risk values ​​of the target ship and other ships based on the position sequence at time points according to an embodiment of this application is shown. Figure 4A flowchart illustrating the method for selecting a target decision scheme from the candidate decision schemes according to an embodiment of this application is shown. Figure 5 A schematic diagram of the structure of the intelligent collision avoidance device for ships according to an embodiment of this application is shown; Figure 6 A schematic diagram of the structure of the electronic device described in an embodiment of this application is shown. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. It should be understood that the accompanying drawings in this application are for illustrative and descriptive purposes only and are not intended to limit the scope of protection of this application. Furthermore, it should be understood that the schematic drawings are not drawn to scale. The flowcharts used in this application illustrate operations implemented according to some embodiments of this application. It should be understood that the operations in the flowcharts may not be implemented in sequence, and steps without logical contextual relationships may be reversed or implemented simultaneously. In addition, those skilled in the art, guided by the content of this application, may add one or more other operations to the flowcharts, or remove one or more operations from the flowcharts.

[0019] Furthermore, the described embodiments are merely some, not all, of the embodiments of this application. The components of the embodiments of this application described and illustrated herein can typically be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely to illustrate selected embodiments of the application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0020] The current mainstream ship collision avoidance technologies mainly suffer from the following problems: the decision-making process of most AI-based collision avoidance systems (such as deep learning models) is not transparent, making it difficult for crew members to understand and trust the actions recommended by the system; traditional systems usually only provide a single "optimal" solution, lacking alternative solutions for comparison, which limits the decision-making flexibility of operators; and collision avoidance systems with fixed parameters are difficult to adapt to changes in different ship characteristics and navigation environments, and lack configurability.

[0021] Moreover, complex optimization algorithms involve huge computational demands in complex scenarios involving multiple ships meeting, making it difficult to meet the timeliness requirements of real-time collision avoidance decision-making; especially in high-density traffic waters, computational delays may lead to missing the best collision avoidance opportunity.

[0022] Based on this, this application provides a ship intelligent collision avoidance method, wherein the method configures a multi-stage behavioral state machine for the target ship; the multi-stage behavioral state machine includes a normal navigation state, a collision avoidance state, a range recovery state, and a cautious navigation state; different states of the state machine correspond to different navigation strategies; when it is determined that there is a collision risk between the target ship and other ships within a preset perception range, the multi-stage behavioral state machine switches to the collision avoidance state, constructing a two-dimensional parameter space corresponding to the target ship; the two-dimensional parameter space includes multiple candidate decision schemes, each candidate decision scheme being a combination of acceleration parameters and turning rate parameters; for each candidate decision scheme in the two-dimensional parameter space, the position sequence of the target ship and other ships within a preset time period in the future is predicted; the other ships are vessels within the preset perception range of the target ship; for each candidate decision scheme, based on The collision risk values ​​of the target ship and other ships are calculated point-in-time according to the position sequence. Based on the collision risk value corresponding to each candidate decision scheme, the target decision scheme is selected from the candidate decision schemes. Different states of the state machine of the method correspond to different navigation strategies. The full search of the two-dimensional parameter space is only initiated in the collision avoidance state. In the normal navigation state, no candidate scheme calculation is required, which avoids the waste of computing resources caused by a single decision process running in the entire time domain. This enables the system to run in the most suitable computing mode in different navigation scenarios, which is particularly suitable for shipborne embedded environments with limited computing resources, while reducing system power consumption. The full search ensures that the global optimal solution can be found. By calculating the collision risk value point-in-time, the entire process of risk evolution over time is captured. It is adapted to high-density traffic waters, the decision-making process is transparent, and a wide variety of solutions are provided.

[0023] The intelligent collision avoidance method for ships is applied to the intelligent control system of ships, hereinafter referred to as the system.

[0024] Please refer to Figure 1 , Figure 1 A flowchart of the intelligent collision avoidance method for ships according to an embodiment of this application is shown; as follows: Figure 1 As shown, the intelligent collision avoidance method for ships includes the following steps S101-S106: S101. Configure a multi-stage behavioral state machine for the target ship; the multi-stage behavioral state machine includes a normal navigation state, a collision avoidance state, a range recovery state, and a cautious navigation state; different states of the state machine correspond to different navigation strategies. S102. When it is determined that there is a collision risk between the target ship and other ships within the preset perception range, the multi-stage behavior state machine switches to the collision avoidance state and constructs a two-dimensional parameter space corresponding to the target ship. The two-dimensional parameter space includes a variety of candidate decision schemes, and each candidate decision scheme is a combination of acceleration parameters and turning rate parameters. S103. For each candidate decision scheme in the two-dimensional parameter space, predict the position sequence of the target ship and other ships in a future preset time period; the other ships are ships within the preset perception range of the target ship. S104. For each candidate decision scheme, calculate the collision risk value of the target ship and other ships point by point in time based on the position sequence; S105. Based on the collision risk value corresponding to each candidate decision scheme, select the target decision scheme from the candidate decision schemes.

[0025] In step S101, a multi-stage behavior state machine for the target ship is configured; the multi-stage behavior state machine includes a normal navigation state, a collision avoidance state, a range recovery state, and a cautious navigation state; different states of the state machine correspond to different navigation strategies.

[0026] The multi-stage behavioral state machine is used to model the ship collision avoidance process as a model with clear state definitions and transition rules.

[0027] The multi-stage behavioral state machine defines four core behavioral states, and the navigation strategy corresponding to each state is as follows: Normal Navigation: When no target vessel enters the collision avoidance zone, the vessel maintains its original course and speed along the planned route; at this time, the output decision parameters cos=0 and rot=0 indicate that no maneuvering action is performed.

[0028] In other words, under normal navigation conditions, the navigation strategy is to output a heading and speed maintenance command, in which the acceleration parameter and the turning rate parameter are both zero.

[0029] Based on this, in some embodiments, the intelligent collision avoidance method for ships further includes: When the trajectory prediction results of the target ship and other ships within the preset perception range are safe within a preset time period in the future, the multi-stage behavior state machine is controlled to be in normal navigation state. When the state machine is in normal navigation state, it determines the normal navigation instruction based on the current heading, current speed and target planned route.

[0030] The normal navigation command is the heading and speed maintenance command.

[0031] In the collision avoidance state, the navigation strategy is to activate a grid search decision algorithm, traverse different combinations of acceleration parameters and turning rate parameters, calculate the risk value of each combination, and finally determine the combination of acceleration parameters and turning rate parameters that meets the preset risk conditions for collision risk value.

[0032] In the course recovery state, the navigation strategy generates a sequence of return instructions based on the ship's current position and planned route.

[0033] In the aforementioned cautious navigation state, the navigation strategy is to reduce speed and pass through.

[0034] The multi-stage behavioral state machine also needs to be configured with state transition rules, which are the transition conditions (or triggering conditions) for each state.

[0035] The transition conditions from normal navigation to collision avoidance, that is, the triggering conditions for the collision avoidance state, are: there is a collision risk between the target ship and other ships within its preset perception range; for example, other ships enter the collision avoidance zone (default 10 nautical miles) and / or the evaluation results of the grid search decision algorithm show that there is a collision risk.

[0036] The transition condition from collision avoidance state to range recovery state is as follows: when the trajectory prediction results of the target ship and other ships within the preset perception range are safe in the future preset time period, such as all target ships have passed the nearest meeting point, that is, the grid search decision algorithm evaluation results show that the collision risk has been eliminated, the system automatically calculates the optimal recovery course.

[0037] The transition from the voyage recovery status to the normal navigation status is determined when the vessel has returned to the planned route (distance from the route is less than the threshold) and the system determines that the voyage recovery is complete.

[0038] Based on this, in some embodiments, the intelligent collision avoidance method for ships further includes: The two-dimensional parameter space is updated in real time, as are the collision risk values ​​of the target ship and other ships. When the collision risk value between the target ship and other ships meets the preset risk elimination conditions, the state machine is controlled to switch to the range recovery state.

[0039] When the state machine is in the range recovery state, a return command is generated based on the projected azimuth angle from the current position of the target ship to the target planned route.

[0040] The transitions between states in the multi-stage behavioral state machine described in this application are smooth and orderly, avoiding sudden changes in decision-making behavior and meeting the requirements of maritime practice.

[0041] If a complex situation or incomplete data is encountered in any of the normal navigation, collision avoidance, or range recovery states, a cautious navigation state will be triggered.

[0042] The target vessel and other vessels within its preset sensing range pose a collision risk. In some embodiments, this risk can be determined based on the following methods: The system acquires the real-time position information of all other ships within a preset sensing range and calculates the real-time distance between the target ship and each other ship. The preset sensing range is configured as a circular area centered on the target ship with a default radius of 10 nautical miles (configurable according to the navigation area and ship characteristics). When the real-time distance between the target ship and any other ship is less than the radius of the preset sensing range, it is determined that the other ship has entered the collision avoidance zone, triggering a preliminary risk warning.

[0043] In some embodiments, triggering a preliminary risk warning can switch to a collision avoidance state.

[0044] In some embodiments, after triggering an initial risk alarm, the system starts a grid search decision algorithm to accurately determine whether there is a collision risk, and this determination is more accurate.

[0045] It should be noted that a grid search decision algorithm can be activated here to determine whether there is a collision risk and whether to switch to a collision avoidance state. After the state machine switches to the collision avoidance state, the grid search decision algorithm is executed to determine the target decision scheme. In some embodiments, the grid search decision algorithm is executed at a certain frequency, such as once every 10 minutes, to determine the most suitable navigation decision scheme in real time.

[0046] In other words, the grid search decision algorithm here can be used to determine whether there is a collision risk between the target ship and other ships within the preset perception range, and it can also be used to determine the target decision scheme in the collision avoidance state.

[0047] The trigger condition for the transition from collision avoidance state to range recovery state is: the trajectory prediction results of the target ship and all other ships within the preset perception range in the future preset time period are safe, and the collision danger has been eliminated.

[0048] In some embodiments, if the trajectory prediction results of the target ship and all other ships within the preset perception range are safe within a preset time period in the future, and the collision risk has been eliminated, specifically, the system re-executes a complete grid search decision algorithm for evaluation. If the quantified risk value of all candidate decision schemes is zero, that is, under all the maneuver strategies that the ship can take, no dangerous state will occur within the preset time period in the future, and the collision risk has been eliminated.

[0049] In some embodiments, the method for determining "the vessel has returned to its planned route" is as follows: The system abstracts the planned route as a broken line formed by connecting multiple waypoints in sequence. During the voyage recovery process, the system continuously calculates the shortest vertical distance (i.e., yaw distance) from the ship's current position to the broken line.

[0050] The method for calculating the shortest vertical distance is as follows: traverse each pair of adjacent waypoints on the planned route, calculate the vertical distance from the ship's current position to each waypoint, and take the minimum value as the yaw distance.

[0051] When the yaw distance is less than the preset recovery threshold (e.g., 0.1 nautical miles, which can be configured according to navigation accuracy requirements and ship handling characteristics), it is determined that the ship has returned to the planned route and meets the conditions for transitioning from the voyage recovery state to the normal navigation state. The cautious navigation state is an alternative state. When encountering complex situations or incomplete data, it can be triggered by any one of the normal navigation state, collision avoidance state, or range recovery state to switch to the cautious navigation state.

[0052] In some embodiments, the conditions that trigger a state of cautious navigation include at least one of the following: The system continuously receives dynamic information messages broadcast by all other ships within the preset sensing range. When a message is missing (e.g., no new message is sent for more than three update cycles) or is abnormal (values ​​exceeding a reasonable range), a cautious navigation state is triggered.

[0053] The system counts the total number of target ships within the preset perception range. When this number exceeds the preset complexity threshold (default 5 ships, which can be configured according to the navigation waters and system computing capabilities), it determines that the current situation is too complex and triggers a cautious navigation state.

[0054] When the conditions that trigger the cautious navigation state are eliminated (such as dynamic information messages returning to normal, the number of other ships dropping below the threshold, sensor data returning to consistency, etc.), the system exits the cautious navigation state.

[0055] In this embodiment, the multi-stage behavioral state machine includes three mutually exclusive main states (normal navigation state, collision avoidance state, and range recovery state) and one superimposed state (cautious navigation state). This allows the system to avoid having to choose between "continuing the current behavior" and "switching to conservative mode" when encountering data anomalies or complex situations. Instead, it can apply safety constraints while maintaining the current behavioral framework, ensuring the continuity of collision avoidance decisions, avoiding decision jumps caused by state switching, and effectively addressing safety requirements when data is incomplete or the environment is complex.

[0056] Specifically, the system operates under a combination of normal navigation and cautious navigation: When the target ship is in normal navigation but encounters incomplete data or a complex situation, the system directly outputs a deceleration command to reduce the ship's speed while maintaining its original course. The grid search decision algorithm is not executed at this time; normal navigation resumes only after the data is recovered or the situation becomes clearer.

[0057] Collision avoidance state + cautious navigation state: When the ship is in a collision avoidance state and the data is incomplete, the grid search decision algorithm is executed. In some embodiments, the number of candidate solutions can be reduced and the calculation speed can be accelerated by appropriately increasing the search step size, so as to adapt to the need for rapid decision-making under data uncertainty.

[0058] Voyage recovery status + Cautious navigation status: When the vessel is in the voyage recovery status and the data is incomplete, maintain a low speed to return to port.

[0059] In step S102, when it is determined that there is a collision risk between the target ship and other ships within the preset perception range, the multi-stage behavior state machine switches to the collision avoidance state and constructs a two-dimensional parameter space corresponding to the target ship; the two-dimensional parameter space includes a variety of candidate decision schemes, and each candidate decision scheme is a combination of acceleration parameters and turning rate parameters; In other words, when it is determined that there is a risk of collision between the target ship and other ships within the preset perception range, the multi-stage behavior state machine switches to the collision avoidance state and executes the grid search decision algorithm.

[0060] The core idea of ​​the grid search decision algorithm is to transform the ship collision avoidance decision problem into a system search problem in a two-dimensional parameter space.

[0061] Please refer to Figure 2 , Figure 2 A flowchart illustrating the method for constructing the two-dimensional parameter space corresponding to the target ship according to an embodiment of this application is shown; the construction of the two-dimensional parameter space corresponding to the target ship includes the following steps S201-S203: S201. Discretize the acceleration parameters with a preset acceleration step size to obtain multiple acceleration sample values; S202. Discretize the head-turn rate parameter with a preset head-turn rate step size to obtain multiple head-turn rate sample values; S203. Combine the acceleration sample value and the head rotation rate sample value in pairs to generate a candidate scheme set containing all combinations.

[0062] The two-dimensional parameter space defines two key decision parameters: acceleration (cos) and turning rate (rot). The acceleration parameter represents the rate of change of the ship's speed, with positive values ​​indicating acceleration and negative values ​​indicating deceleration, and the unit is knots per minute; the turning rate parameter represents the rate of change of the ship's course, with positive values ​​indicating a right turn and negative values ​​indicating a left turn, and the unit is degrees per minute.

[0063] During the parameter space construction phase, the system automatically generates discretized parameter vectors based on the configured parameter range and search step size (preset acceleration step size, preset rotation speed step size).

[0064] For example, the preset acceleration step size is 0.1, and the acceleration parameter space is {0, ±0.1, ±0.2, ±0.3, ±0.4, ±0.5}, with a total of 11 sampling points; the preset turning rate step size is 1, and the turning rate parameter space is {0, ±1, ±2, ±3, ±4, ±5, ±6, ±7, ±8}, with a total of 17 sampling points. The Cartesian product of the two parameter spaces forms 187 combinations, thus obtaining a candidate solution set including 187 candidate decision schemes.

[0065] The two-dimensional parameter space design has the following characteristics: comprehensive coverage, with 187 candidate schemes that can fully represent the possible maneuvers of the ship; high search efficiency, with fixed-step grid division avoiding redundant calculations; and strong configurability, allowing users to adjust the parameter range according to ship characteristics, water conditions, and real-time requirements.

[0066] The candidate decision scheme can also be referred to as the candidate scheme.

[0067] In step S103, for each candidate decision scheme in the two-dimensional parameter space, the position sequence of the target ship and other ships in the future within a preset time period is predicted.

[0068] For each candidate decision, the system executes a complete trajectory prediction and risk assessment process. First, based on kinematic equations, it predicts the ship's position sequence over a preset time period (e.g., within the next 600 seconds). Specifically, the initial state is determined by the ship's current position, heading, and speed. The acceleration parameter is used as the rate of change of speed, and the turning rate parameter is used as the rate of change of heading. Within each prediction time step (e.g., 10 seconds), the model calculates the speed for the next time step based on the current speed and acceleration, and the heading for the next time step based on the current heading and turning rate. Then, the position coordinates for the next time step are deduced based on the speed, heading, and time step length. This iterative process is repeated step by step until a preset prediction duration (600 seconds) is reached, generating a complete prediction trajectory containing 60 discrete trajectory points.

[0069] At the same time, the system also predicts the position sequence of all other ships that have entered the preset sensing range within the next 600 seconds, based on their current position, heading and speed (usually obtained through AIS, and it is assumed that they will maintain their current motion state during the prediction period, i.e., maintain heading and speed).

[0070] By aligning the predicted trajectory of the ship under each candidate decision with the predicted trajectories of all other ships in time and space, the system can accurately determine the relative positional relationship between the ship and all other ships at any future time under each candidate decision, providing a data foundation for subsequent point-in-time collision risk assessment.

[0071] Determine the ship's position sequence over the next 600 seconds; then, for each predicted time point, calculate the distance between the ship and all other ships; finally, count the number of dangerous situations and the time of the first danger.

[0072] In step S104, for each candidate decision scheme, the collision risk value of the target ship and other ships is calculated point-in-time based on the position sequence; Please refer to Figure 3 The calculation of the collision risk value between the target ship and other ships based on the position sequence at each time point includes the following steps S301-S303: S301. For each time point in the position sequence, calculate the distance between this ship and other ships; S302. If the distance is less than a preset distance threshold, then mark the time point as a dangerous state between the target ship and the other ship. S303. Accumulate the number of dangerous situations the target ship experiences at all time points and among all other ships to obtain the collision risk value for each candidate decision scheme.

[0073] Here, for the predicted position sequences of the ship and other ships under a certain candidate decision scheme, the system calculates the spatial distance between the ship and each other ship at each of the 60 predicted time points. If the distance between the ship and another ship at a certain time point is less than a preset safe encounter distance threshold (3 nautical miles by default), the system immediately marks that time point as a dangerous state. After this process traverses all time points and all other ships, the system uses the total number of dangerous states accumulated as the collision risk value of the candidate decision scheme.

[0074] In other words, the danger assessment uses a distance threshold method: when the distance between two ships is less than the safe meeting distance (default 3 nautical miles, configurable), the time point is considered to be in a dangerous state. The number of dangers (danger_size) is accumulated across all time points and all target ships, serving as a quantitative indicator of the risk level of the plan.

[0075] The collision risk value can also be called the quantified risk value.

[0076] The collision risk value obtained in this way can comprehensively reflect the entire process of risk evolution within the entire prediction time window. For example, two ships may meet the safe distance at the nearest meeting point, but enter the danger distance range multiple times during the approach and departure process; or in a multi-ship situation, one ship may be safe but another ship may approach.

[0077] It should be noted that for scenarios involving multiple target vessels, the system employs a joint assessment strategy. The number of dangerous encounters with all target vessels is summed up to obtain the overall risk quantification value for the plan. This strategy ensures that collision avoidance decisions consider not only a single target vessel but also all encountered targets comprehensively, avoiding situations where avoiding one target leads to new encounters with other targets.

[0078] Traditional DCPA / TCPA methods only assess a single nearest encounter point, which is a qualitative judgment and cannot reflect the dynamic evolution of risk. The method described in this application uses continuous assessment at 60 time points to achieve quantitative quantification of risk. Traditional methods assume that the target ship maintains its heading and speed, which does not match actual navigation. The method described in this application considers the changes in the ship's acceleration and turning rate, and can reflect the maneuvering effect. Traditional methods use independent assessment and serial processing for multiple targets. The method described in this application uses joint assessment and parallel optimization.

[0079] The evaluation results of the candidate decision-making schemes are encapsulated in a data structure. The evaluation results include: decision parameters (cos, rot), quantified risk value (danger_size), and hazard details for each target vessel (including target vessel identification and the time of the first hazard). The evaluation results of all 187 candidate schemes are fully preserved, forming a traceable decision knowledge base.

[0080] In step S105, a target decision scheme is selected from the candidate decision schemes based on the collision risk value corresponding to each candidate decision scheme.

[0081] Please refer to Figure 4 The step of selecting the target decision scheme from the candidate decision schemes based on the collision risk value corresponding to each candidate decision scheme includes the following steps S401-S402: S401. Determine the candidate decision scheme whose collision risk value meets the preset risk conditions as the target decision scheme; S402. When multiple target decision schemes have the same collision risk value, the priority of the target decision schemes is determined based on the steering amplitude.

[0082] For example, the preset risk condition is to select one or more candidate decision schemes with the lowest collision risk value as the target decision scheme.

[0083] When multiple decision-making options are provided, crew members can choose based on their experience and operational capabilities.

[0084] In some embodiments, candidate solutions are sorted in ascending order of risk quantification value, i.e., the smaller the danger_size, the higher the priority. When multiple solutions have the same danger_size, they are sorted in ascending order of absolute turning angle value, i.e., the solution with the smallest turning angle is selected. This sorting criterion reflects the design philosophy of "safety first, with economic considerations in between".

[0085] The system always selects the optimally ranked solution as the output, while retaining complete information on all candidate solutions. This design allows operators not only to obtain the optimal decision instruction but also to understand the entire decision space and verify the rationality of the system's decisions.

[0086] In some embodiments, the intelligent collision avoidance method for ships further includes: Record each candidate decision scheme in the candidate scheme set and the corresponding collision risk value, as well as the feedback information during the execution of the target decision scheme.

[0087] In this embodiment, after performing a grid search and completing a collision risk assessment of all candidate decision schemes, the system does not only save the final selected target decision scheme, but also persistently stores all the complete data from this collision avoidance decision process. Specifically, for each candidate decision scheme generated in the two-dimensional parameter space, the system records the following information: Decision parameter information, including the acceleration parameter value and the turning speed parameter value corresponding to the candidate decision scheme.

[0088] The assessment results include the collision risk value (i.e., the number of dangerous situations) corresponding to the candidate decision-making scheme, as well as intermediate data from the assessment process. The intermediate data includes: the time point at which each other vessel first experiences a dangerous situation under this scheme (i.e., the time when the distance between this vessel and that vessel first falls below the safety threshold), the distribution sequence of all dangerous situation time points, and a detailed list of collision risks for all other vessels under this scheme. This intermediate data allows operators to gain a deeper understanding of the risk distribution for each scheme. For example, two candidate schemes may have the same total number of dangerous situations, but one scheme's dangerous situations may be concentrated in the early part of the prediction time window, while the other's are concentrated in the later part. The risk characteristics of the two are fundamentally different. The intermediate data fully preserves these differences, providing a data foundation for operators' auxiliary judgment and post-event analysis.

[0089] After selecting the target decision scheme, the system records the execution information of the scheme in its entirety. The execution information includes decision output information, such as the acceleration parameters and turning rate parameters corresponding to the target decision scheme, the decision timestamp, the ship's position / heading / speed at the time of the decision, and the collision risk value of the target decision scheme; it also includes actual execution result information, such as the ship's actual position sequence, heading change sequence, and speed change sequence after executing the target decision scheme, thereby recording the complete motion process from the moment the decision is executed until the decision effect is fully manifested (i.e., the ship reaches a new stable navigation state).

[0090] The solution described in this application, by fully recording each candidate solution and its evaluation results, allows operators not only to know "what the system selected," but also to understand the full picture of the decision space, such as which options were considered, why these options were excluded, and why this option was selected. This transforms collision avoidance decision-making from a black box to a white box, significantly enhancing the crew's trust in the system.

[0091] With complete recording, each collision avoidance decision-making task forms a complete data asset, which can be used for accident replay analysis, crew training and teaching, and continuous optimization and iteration of system algorithms.

[0092] Based on the same inventive concept, this application also provides a ship intelligent collision avoidance device corresponding to the ship intelligent collision avoidance method. Since the principle of the device in this application is similar to that of the ship intelligent collision avoidance method described above in this application, the implementation of the device can refer to the implementation of the method, and the repeated parts will not be described again.

[0093] Please refer to Figure 5 , Figure 5 This paper shows a schematic diagram of the structure of a ship intelligent collision avoidance device according to an embodiment of this application. The device includes: Configuration module 501 is used to configure the multi-stage behavioral state machine of the target ship; the multi-stage behavioral state machine includes normal navigation state, collision avoidance state, range recovery state and cautious navigation state; different states of the state machine correspond to different navigation strategies. The construction module 502 is used to construct a two-dimensional parameter space corresponding to the target ship when it is determined that there is a collision risk between the target ship and other ships within the preset perception range. The two-dimensional parameter space includes a variety of candidate decision schemes, and each candidate decision scheme is a combination of acceleration parameters and turning rate parameters. Prediction module 503 is used to predict the position sequence of the target ship and other ships in a future preset time period for each candidate decision scheme in the two-dimensional parameter space; The calculation module 504 is used to calculate the collision risk value of the target ship and other ships on a time-by-time basis based on the position sequence for each candidate decision scheme; The filtering module 505 is used to filter out the target decision scheme from the candidate decision schemes based on the collision risk value corresponding to each candidate decision scheme.

[0094] In some embodiments, in the intelligent collision avoidance device for ships, the building module is further configured to: discretize the acceleration parameters with a preset acceleration step size to obtain multiple acceleration sample values; The head-turn rate parameter is discretized and sampled with a preset head-turn rate step size to obtain multiple head-turn rate sample values; The acceleration sample value and the head rotation rate sample value are combined in pairs to generate a candidate scheme set containing all combinations.

[0095] In some embodiments, in the intelligent collision avoidance device for ships, the evaluation module is further configured to: calculate the distance between the ship and other ships for each time point in the position sequence; If the distance is less than a preset distance threshold, then the time point is marked as a dangerous situation between the target ship and the other ship; The collision risk value for each candidate decision scheme is obtained by summing the number of dangerous situations the target ship experiences at all time points and among all other ships.

[0096] In some embodiments, in the intelligent collision avoidance device for ships, the screening module is further configured to: determine candidate decision schemes whose collision risk values ​​meet preset risk conditions as target decision schemes; when multiple target decision schemes have the same collision risk value, determine the priority of the target decision schemes based on the steering amplitude.

[0097] In some embodiments, the intelligent collision avoidance device for ships further includes: The first control module is used to control the multi-stage behavior state machine to a normal navigation state when the trajectory prediction results of the target ship and other ships within the preset perception range are safe in the future preset time period. When the state machine is in normal navigation state, it determines the normal navigation instruction based on the current heading, current speed and target planned route.

[0098] In some embodiments, the intelligent collision avoidance device for ships further includes: The second control module is used to update the two-dimensional parameter space in real time, and to update the collision risk values ​​of the target ship and other ships in real time. When the collision risk value of the target ship and other ships meets the preset risk elimination conditions, the state machine is controlled to switch to the range recovery state; When the state machine is in the range recovery state, a return command is generated based on the projected azimuth angle from the current position of the target ship to the target planned route. In some embodiments, the intelligent collision avoidance device for ships further includes: The recording module is used to record each candidate decision scheme in the candidate scheme set and the collision risk value corresponding to the candidate decision scheme, as well as to record the feedback information during the execution of the target decision scheme.

[0099] Based on the same inventive concept, this application also provides an electronic device corresponding to the intelligent collision avoidance method for ships. Since the principle of solving the problem by the electronic device in this application is similar to that of the intelligent collision avoidance method for ships described above in this application, the implementation of the electronic device can refer to the implementation of the method, and the repeated parts will not be described again.

[0100] Please refer to Figure 6 , Figure 6 A schematic diagram of the structure of the electronic device according to an embodiment of this application is shown. The electronic device 600 includes a processor 601, a memory 602 and a bus. The memory 602 stores machine-readable instructions that can be executed by the processor 601. When the electronic device 600 is running, the processor 601 and the memory 602 communicate through the bus. When the machine-readable instructions are executed by the processor 601, the steps of the intelligent collision avoidance method for ships are performed.

[0101] Based on the same inventive concept, this application also provides a computer-readable storage medium corresponding to the intelligent collision avoidance method for ships. Since the principle of the computer-readable storage medium in this application is similar to that of the intelligent collision avoidance method for ships described above, the implementation of the computer-readable storage medium can refer to the implementation of the method, and the repeated parts will not be described again.

[0102] This application also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, performs the steps of the aforementioned intelligent collision avoidance method for ships.

[0103] It should be noted that the term "comprising" will be used in the embodiments of this application to indicate the presence of the features declared thereafter, but does not exclude the addition of other features.

[0104] In the embodiments provided in this application, it should be understood that the disclosed apparatus and methods can be implemented in other ways. The apparatus embodiments described above are merely illustrative. For example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. Furthermore, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Additionally, the displayed or discussed mutual couplings, direct couplings, or communication connections may be through some communication interfaces; indirect couplings or communication connections between devices or units may be electrical, mechanical, or other forms.

[0105] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0106] In addition, the functional units in the embodiments provided in this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0107] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0108] It should be noted that similar labels and letters in the following figures indicate similar items. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. In addition, the terms "first", "second", "third", etc. are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0109] Finally, it should be noted that the above-described embodiments are merely specific implementations of this application, used to illustrate the technical solutions of this application, and not to limit them. The protection scope of this application is not limited thereto. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that any person skilled in the art can still modify or easily conceive of changes to the technical solutions described in the foregoing embodiments, or make equivalent substitutions for some of the technical features, within the scope of the technology disclosed in this application; and these modifications, changes, or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application. All should be covered within the protection scope of this application. Therefore, the protection scope of this application should be determined by the protection scope of the claims.

Claims

1. A ship intelligent collision avoidance method, characterized in that, The method includes: Configure a multi-stage behavioral state machine for the target ship; the multi-stage behavioral state machine includes a normal navigation state, a collision avoidance state, a range recovery state, and a cautious navigation state; different states of the state machine correspond to different navigation strategies. When it is determined that there is a collision risk between the target ship and other ships within the preset perception range, the multi-stage behavioral state machine switches to the collision avoidance state and constructs a two-dimensional parameter space corresponding to the target ship. The two-dimensional parameter space includes a variety of candidate decision schemes, and each candidate decision scheme is a combination of acceleration parameters and turning rate parameters. For each candidate decision scheme in the two-dimensional parameter space, predict the position sequence of the target ship and other ships in the future within a preset time period; For each candidate decision scheme, the collision risk value of the target ship and other ships is calculated point by point in time based on the position sequence; Based on the collision risk value corresponding to each candidate decision scheme, the target decision scheme is selected from the candidate decision schemes.

2. The intelligent collision avoidance method for ships according to claim 1, characterized in that, The construction of the two-dimensional parameter space corresponding to the target ship includes: The acceleration parameters are discretized and sampled with a preset acceleration step size to obtain multiple acceleration sample values; The head-turn rate parameter is discretized and sampled with a preset head-turn rate step size to obtain multiple head-turn rate sample values; The acceleration sample value and the head rotation rate sample value are combined in pairs to generate a candidate scheme set containing all combinations.

3. The intelligent collision avoidance method for ships according to claim 1, characterized in that, Based on the location sequence, the collision risk values ​​of the target ship and other ships are calculated point-in-time, including: For each time point in the position sequence, calculate the distance between this ship and other ships; If the distance is less than a preset distance threshold, then the time point is marked as a dangerous situation between the target ship and the other ship; The collision risk value for each candidate decision scheme is obtained by summing the number of dangerous situations the target ship experiences at all time points and among all other ships.

4. The intelligent collision avoidance method for ships according to claim 1, characterized in that, The step of selecting the target decision scheme from the candidate decision schemes based on the collision risk value corresponding to each candidate decision scheme includes: Candidate decision schemes whose collision risk values ​​meet preset risk conditions are identified as target decision schemes; When multiple target decision options have the same collision risk value, the priority of the target decision options is determined based on the steering magnitude.

5. The intelligent collision avoidance method for ships according to claim 1, characterized in that, The method further includes: When the trajectory prediction results of the target ship and other ships within the preset perception range are safe within a preset time period in the future, the multi-stage behavior state machine is controlled to be in normal navigation state. When the state machine is in normal navigation state, it determines the normal navigation instruction based on the current heading, current speed and target planned route.

6. The intelligent collision avoidance method for ships according to claim 1, characterized in that, The method further includes: The two-dimensional parameter space is updated in real time, as are the collision risk values ​​of the target ship and other ships. When the collision risk value between the target ship and other ships meets the preset risk elimination conditions, the state machine is controlled to switch to the range recovery state. When the state machine is in the range recovery state, a return command is generated based on the projected azimuth angle from the current position of the target ship to the target planned route.

7. The intelligent collision avoidance method for ships according to claim 2, characterized in that, The method further includes: Record each candidate decision scheme in the candidate scheme set and the corresponding collision risk value, as well as the feedback information during the execution of the target decision scheme.

8. A ship intelligent collision avoidance device, characterized in that: The device includes: The configuration module is used to configure the multi-stage behavior state machine of the target ship; the multi-stage behavior state machine includes normal navigation state, collision avoidance state, range recovery state, and cautious navigation state; different states of the state machine correspond to different navigation strategies; The construction module is used to construct a two-dimensional parameter space corresponding to the target ship when it is determined that there is a collision risk between the target ship and other ships within the preset perception range. The two-dimensional parameter space includes multiple candidate decision schemes, and each candidate decision scheme is a combination of acceleration parameters and turning rate parameters. The prediction module is used to predict the position sequence of the target ship and other ships in a future preset time period for each candidate decision scheme in the two-dimensional parameter space. The calculation module is used to calculate the collision risk value of the target ship and other ships on a time-by-time basis based on the position sequence for each candidate decision scheme; The filtering module is used to filter out the target decision scheme from the candidate decision schemes based on the collision risk value corresponding to each candidate decision scheme.

9. An electronic device, characterized in that, include: The device includes a processor, a memory, and a bus. The memory stores machine-readable instructions executable by the processor. When the electronic device is running, the processor communicates with the memory via the bus. When the machine-readable instructions are executed by the processor, they perform the steps of the ship intelligent collision avoidance method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, performs the steps of the intelligent collision avoidance method for ships as described in any one of claims 1 to 7.

Images