An intelligent ship autonomous collision avoidance whole-process path planning method

Abstract

The application discloses a kind of intelligent ship autonomous collision avoidance whole process path planning method, comprising the following steps: according to the dynamic information and static information of target ship, establish the error distribution model of fusion target, analysis obtains the classification model of target ship;According to the classification model of target ship, combine four ship field model, establish the ship field model of dynamic and static obstacle;Local path planning is established based on artificial potential field method, repulsion field is established around the obstacle in ship field model, gravity field is established at target point, global path return mechanism is introduced, and local collision avoidance path planning algorithm considering return mechanism is established;Based on the degree of current ship path deviation, navigation deviation, realize gravity self-adaptive adjustment, form dynamic optimization local collision avoidance path planning algorithm considering adaptive return mechanism and carry out the path planning of ship. After the ship collision avoidance of the application, control ship to return to global route reasonably and efficiently, ensure the safety and efficiency of ship navigation.

Description

Technical Field

[0001] The present invention belongs to the technical field of waterway transportation, and in particular relates to a full-process path planning method for autonomous collision avoidance of intelligent ships. Background Art

[0002] In recent years, with the increasing domestic and international trade, the shipping industry has experienced unprecedented development. The number of ships has increased dramatically, the size of the fleet has continued to expand, and the density of ship traffic has also increased. This is particularly true in coastal waterways and near ports, where ships are highly concentrated and maritime collisions are a frequent occurrence, causing significant economic losses and safety threats to the shipping industry. Research has found that the root cause of most collisions is human error, with crew misjudgment, fatigue, inattention, and operational errors being the primary contributing factors. Research on autonomous collision avoidance has become a key approach to addressing this issue and holds significant practical significance. Intelligent ships possess strong autonomous decision-making and control capabilities, enabling them to rapidly generate collision avoidance strategies in complex navigation environments. They are an inevitable trend in the future of safe and green ship development. However, traditional intelligent ships focus more on the effectiveness of collision avoidance decisions and path planning, while paying less attention to the return process after autonomous collision avoidance. In fact, after clearing another ship, they often deviate significantly from their path or make large turns, making it difficult for the ship to safely and properly return to its overall route, increasing energy consumption and the risk of accidents. Summary of the Invention

[0003] The purpose of the present invention is to provide a full-process path planning method for autonomous collision avoidance of intelligent ships, which controls the ship to return to the global route reasonably and efficiently after the ship avoids the collision, thereby ensuring the safety and efficiency of the ship's navigation.

[0004] To solve the above technical problems, the technical solution of the present invention is: a full-process path planning method for autonomous collision avoidance of intelligent ships, comprising the following steps:

[0005] S1. Obtain dynamic and static information of the target ship through navigation radar and automatic identification system, establish an error distribution model of the fused target based on the dynamic and static information of the target ship, and obtain a classification model of the target ship based on the error distribution model of the fused target;

[0006] S2. Based on the classification model of the target ship and the four-element ship domain model, a ship domain model of dynamic and static obstacles is established;

[0007] S3. Establish local path planning based on the artificial potential field method. A repulsive field is established around obstacles in the ship domain model, and a gravitational field is established at the target point. At the same time, a global path return mechanism is introduced, and a gravitational field is established at the nodes of the global path. A local collision avoidance path planning algorithm is established that takes the return mechanism into account.

[0008] S4. Adaptive gravity adjustment is implemented based on the current ship path deviation and the degree of navigation deviation. The dynamic ship domain model and repulsive field are adjusted based on the relative motion state of the obstacle. Given a time step, the local path planning algorithm is corrected according to the error of each planning and execution, forming a dynamically optimized local collision avoidance path planning algorithm that takes into account the adaptive return mechanism. The ship's path is planned based on this local collision avoidance path planning algorithm.

[0009] The dynamic information of the target ship in S1 includes at least the real-time distance, speed, heading, position, longitude and latitude, and time of the target ship, and the static information includes at least the ship name, water mobile service identification code, and navigation status of the target ship.

[0010] S1 is specifically:

[0011] Collecting raw data from each vessel using marine radar and automatic identification systems. The raw data includes at least speed, heading, longitude and latitude, and time. Generating the historical trajectory of the vessel based on the time sequence and longitude and latitude changes in the raw data. Collecting salient features of the historical trajectory data. The salient features include at least the rate of change of speed, rate of change of heading, and position change.

[0012] Perform data cleaning on prominent features, check for missing values ​​and abnormal data, interpolate and fill missing values ​​based on historical trajectories, remove abnormal data, and ensure the integrity of the ship's historical trajectory data;

[0013] Set the parameters of the OPTICS clustering model, including the minimum number of neighbors and the maximum neighborhood distance, and develop a metric;

[0014] The normalized historical trajectory data is used as input to the OPTICS clustering model. The OPTICS clustering model is executed to automatically identify and sort the density reachability of data points, and output a variable-density clustering result. The clustering result includes at least the reachability distance and its sorting graph, as well as cluster labels. The cluster division is determined based on the reachability distance sorting graph, and the clusters are classified based on the output cluster labels.

[0015] Based on the clustering results and the static information provided by the automatic identification system, the ship classification decision is made and the classification model of moving ships and anchored ships is constructed;

[0016] If the ship trajectory of a cluster in the clustering result is similar to that of a moored ship, and the static information of the automatic identification system indicates that the ship is in a stationary or moored state, then the cluster is marked as a moored ship; if the ship trajectory of a cluster is similar to that of a moving ship, and its static information indicates that the ship is in a moving state, then the cluster is marked as a moving ship. The four-element ship domain model QSD in S2 is expressed as:

[0017] QSD={(x,y)∣f(x,y;Q)≤1,Q={R fore ,R aft ,R starb ,R port}}

[0018]

[0019] Among them, f(x, y; Q) is the function that determines the boundary of the four-dimensional ship domain, x and y are the spatial coordinates of the ship, Q is an intermediate variable, representing the four radii of the ship domain, sgn(x) and sgn(y) represent a sign function; R fort 、R aft 、R starb 、R port Respectively represent the radius of the four-element ship field in the front, back, right and left directions;

[0020] The quaternary ship domain is elliptical in shape, and its domain size is determined by R fort 、R aft 、R starb 、R port Four radius parameters are determined, and the boundary shape of the quaternion ship field is adjusted by these four radius parameters to cope with different heading scenarios.

[0021] The method of establishing the ship domain model for dynamic and static obstacles is as follows: according to the classification model of target ships, the four-element ship models of different types of target ships are adjusted to establish static and dynamic ship domain models respectively;

[0022] If the target ship is a moving ship, it can be regarded as a dynamic obstacle, that is, a dynamic ship domain model is established for it, and the values ​​of its four radius parameters are:

[0023]

[0024] Where L is the length of the ship, k AD k is the gain coefficient of the ship's distance AD ​​in the circle, DT is the gain coefficient of the initial diameter of the cycle DT;

[0025] If the target ship is a moored ship, it can be regarded as a static obstacle, that is, a static ship domain model is established for it, and the values ​​of its four radius parameters are:

[0026]

[0027] S3 specifically:

[0028] According to the artificial potential field method, a repulsive field is established around the obstacle, the field is scaled and adjusted according to the radius parameters of dynamic and static obstacles, a gravitational field is constructed around the target point, and a return gravitational field is established according to the path nodes;

[0029] The potential field of a ship is changing at every moment. Assume that at a certain moment, there are N s static obstacles, N d dynamic obstacles, the current sum of the artificial potential field is the sum of the gravitational potential field and the repulsive potential field, expressed as:

[0030]

[0031] Among them, P f is the sum of the current artificial potential field, P att is the gravitational field, P ref is the repulsive field, P g is the target point potential field, P n is the gravitational field of the global path node, P s,i is the potential field of point obstacle i, P d,j is the potential field of domain-type obstacle j;

[0032] During the local collision avoidance process, the collision risk index (CRI) is used to determine the time points for the ship to start collision avoidance and return. When the CRI is higher than the first collision threshold, the ship begins collision avoidance. By combining the current artificial potential field sum with map modeling, the minimum potential field near the ship is found. The ship will tend to move towards the area with smaller potential field. As the collision avoidance process is completed, the collision risk decreases. When the CRI is lower than the second collision threshold, the time to return is confirmed, thus realizing local path planning and autonomous decision-making for obstacle avoidance.

[0033] S4 is specifically:

[0034] After confirming the timing of returning home, deviation detection is performed to calculate the distance between the ship and the global path, as well as the deviation between the ship's heading and the global path heading;

[0035] The formula for the distance that a ship deviates from the global path is expressed as:

[0036] Δd=||(x,y)-(x p ,y p )||

[0037] Among them, Δd is the distance from the ship to the global path, (x, y) is the current position of the ship, (x p ,y p ) is the global path node closest to the ship;

[0038] The formula for the angle at which a ship deviates from its global heading is:

[0039] Δψ=|ψ-ψ0|

[0040] Where Δψ is the deviation between the ship heading and the global path heading, ψ is the current heading angle of the ship, and ψ0 is the global path heading angle;

[0041] To unify the dimensions, normalization is performed, assuming that the maximum path deviation does not exceed d m , the maximum heading deviation does not exceed ψ m ;

[0042] The degree of deviation from the path d d Expressed as:

[0043]

[0044] Degree of deviation from the course d ψ Expressed as:

[0045]

[0046] Dynamically adjust the gravitational potential field based on the deviation error. If there is an obstacle, adjust the dynamic ship field and repulsive potential field based on the relative motion state of the obstacle.

[0047] Establish an adaptive return mechanism, including: parameter and target initialization, adaptive gravitational field adjustment, dynamic ship field and adaptive repulsive field adjustment, adaptive total potential field adjustment and update.

[0048] The parameters and target initialization are as follows: Assume that the speed at the moment of returning after the collision is v1, the heading angle is ψ1, the path deviation is Δd1, the heading deviation is Δψ1, and the initial ship area radius is R fore , R aft , R starb , R port , the minimum encounter distance is DCPA;

[0049] To achieve the return goal, set the path deviation to be less than d min When the heading deviation is less than ψ min When , it is regarded as returning to the global heading;

[0050] The specific steps of adaptive gravitational field adjustment are:

[0051] Set the time step Δt, and update the path deviation d every Δt d , heading deviation d ψ , speed deviation Δv and recalculate the weights of the return gravity field and the target gravity field. The calculation formula is expressed as:

[0052]

[0053] Among them, k1 and k2 are the attenuation degrees. The larger the value, the faster the attenuation.

[0054] The adaptive gravitational field is expressed as:

[0055] P att =(1+ω1)P n +(1+ω 2) P g

[0056] Among them, ω1 and ω2 are the weights of the return gravitational field and the target gravitational field.

[0057] The specific steps for dynamic ship field and adaptive repulsive field adjustment are:

[0058] After the ship avoids collision, the collision risk of the ship is judged according to the relative position, relative speed and relative heading of the ship. In order to better detect the collision risk, the ship area is expanded k times and an external safety distance d is added. s , expressed as:

[0059] d s =k·R

[0060]

[0061] in, Indicates the boundary detection range of the expanded ship area, R is the length from the center of the area to the boundary, and k is the multiple of the expanded range;

[0062] According to different encounter scenarios, it is judged whether there is a collision risk between the own ship and the target ship. A radius of d is constructed with the own ship as the center. s The detection circle is adjusted according to the different scenes;

[0063] A detection line is constructed based on the relative positions and headings of the ships. Whether the detection line passes through the range of the detection circle is used to determine whether there is a collision risk between the two ships. The closest approach distance (DCPA) is one of the main factors in determining the risk of collision. It represents the distance between the two ships when they are in their current state of motion and continue to sail at a predetermined heading and speed. The smaller the DCPA value, the greater the collision risk. The size of the dynamic ship area is adjusted based on speed, heading, and position to enable ships to start collision avoidance in advance. The specific steps are as follows:

[0064] Assuming R1 is the distance from the intersection of the DCPA line and the ship area to the ship, and R2 is the distance from the intersection of the DCPA extension line and the detection circle to the ship, then,

[0065]

[0066] Among them, ω DCPAIndicates the adjustment weight of the ship's field radius; is the attenuation coefficient. The larger its value, the faster the attenuation. DCPA represents the closest approach distance between own ship and target ship.

[0067] Adjust the ship domain based on the direction of the ship encounter; when there is a collision risk, adjust the ship domain model QSD based on the direction of the target ship e , the adjustment formula is expressed as:

[0068] QSD e ={(x,y)|f e (x,y;Q)≤1,

[0069] Q e ={R fore +ΔR fore ,R aft +ΔR aft ,R starb +ΔR starb ,R port +ΔR port}}

[0070] ΔR r =R r ·ω DCPA

[0071] Among them, f e (x, y; Q) represents the boundary of the ship area, Q e Represents the adjustment value of the radius of the four areas, ΔR r Indicates the dynamic adjustment value of the field radius, r indicates the direction of adjustment, ΔR fore Indicates the dynamic adjustment of the forward radius of the ship, ΔR aft Indicates the dynamic adjustment of the ship's rearward radius, ΔR starb Indicates the dynamic adjustment of the starboard radius of the ship, ΔR port Indicates the dynamic adjustment of the port radius of the ship, R r ∈{R fore ,R aft ,R starb ,R port};

[0072] After avoiding collision, return to the destination. For obstacles behind, the return process needs to avoid collision with obstacles behind. Define Δv OT = The speed of the own ship divided by the speed of the obstacle behind it, expressed as:

[0073]

[0074] Among them, v OS Indicates the ship's speed, v TSIndicates the speed of the target ship;

[0075] Δθ OT is the angle between the speed directions of the own ship and the target ship, and the velocity repulsion field P of the dynamic obstacle is introduced. v , which is expressed as:

[0076]

[0077] in, is the relative velocity Δv OT The gain factor, Δv OT The smaller it is, the greater the possibility of the obstacle behind catching up. A larger value should be taken;

[0078] Adaptive repulsive field P of the ship ref for:

[0079]

[0080] The specific steps of adaptive total potential field adjustment and update are:

[0081] After each time step Δt, the current DCPA and relative speed Δv are recalculated. OT and the angle Δθ OT , update the current total adaptive potential field; during the ship's return process, its adaptive total potential field is:

[0082]

[0083] Among them, p att is the current total gravitational field, P ref is the current total repulsive field, P g is the target point potential field, P n is the gravitational field of the global path node, P s,i is the potential field of point obstacle i, P d,j is the potential field of domain-type obstacle j, P v is the velocity repulsion field of the dynamic obstacle, ω1 and ω2 are the update weights of the return gravitational field and the target gravitational field.

[0084] A computer-readable storage medium is also provided, on which a computer program is stored, and when the computer program is executed by a processor, the steps of any of the above methods are implemented.

[0085] Compared with the prior art, the present invention has the following beneficial effects:

[0086] The present invention uses an adaptive return mechanism to enable a ship to return to the global route more efficiently and safely during the return process after collision avoidance, and can adapt to complex offshore navigation environments, thereby improving the intelligence level of the ship's autonomous collision avoidance and return. BRIEF DESCRIPTION OF THE DRAWINGS

[0087] Figure 1 A schematic diagram of a flow chart of an embodiment of the present invention;

[0088] Figure 2 This is a flowchart of navigation data processing in an embodiment of the present invention;

[0089] Figure 3 Schematic diagram of the adaptive return-to-home artificial potential field in an embodiment of the present invention;

[0090] Figure 4 This is a flow chart of the adaptive return-to-home mechanism in an embodiment of the present invention;

[0091] Figure 5 This is a flowchart of the full process of autonomous collision avoidance path planning in an embodiment of the present invention. DETAILED DESCRIPTION

[0092] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely for the purpose of explaining the present invention and are not intended to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below may be combined with each other as long as they do not conflict with each other.

[0093] The technical solution of the present invention is:

[0094] A full-process path planning method for autonomous collision avoidance of intelligent ships, comprising:

[0095] Step 1: Fusion of target ship data from the Automatic Identification System (AIS) and marine radar to establish an error distribution model for the fused target. Target ship information, including the ship's speed, heading, position, and navigation status, is obtained through marine radar and AIS. AIS provides information such as the ship's name, Maritime Mobile Service Identity (MMSI), and navigation status. Marine radar can obtain real-time information such as the target ship's distance, speed, heading, and position. Marine radar's short-range dynamic information is more accurate, while AIS's long-range information performs better. By combining the static information provided by AIS with the dynamic information provided by marine radar, the dynamic information from marine radar is fused with the speed and heading information from AIS to establish an error distribution model for the fused target to improve the positioning accuracy of the target ship.

[0096] Based on the error distribution model of the fusion target, an Ordering Points to Identify the Clustering Structure (OPTICS) algorithm was used to establish a classification model for moored and moving ships. First, raw data for each ship, including speed, heading, latitude and longitude, and time, was collected via AIS and marine radar. The ship's historical trajectory was generated using the time sequence and latitude and longitude changes. Key features of the data, such as speed change rate, heading change rate, and position change, were collected. The data was cleaned, missing values ​​and outliers were checked, and missing values ​​were interpolated based on the historical trajectory data to remove outliers, ensuring the integrity of the ship trajectory data. The OPTICS model parameters, including the minimum number of neighbors and maximum neighborhood distance, were then set, and a metric was developed. The OPTICS clustering model was then executed, taking the normalized dataset as input and using the set parameters. The algorithm automatically identified and ranked the data points by density reachability, outputting a variable-density clustering result, which primarily consisted of reachability distance, ranking, and cluster labels. Cluster division was determined by observing the reachability distance ranking graph, and clusters were classified based on the output cluster labels. Finally, based on the clustering results of OPTICS and the ship information provided by AIS, the ship classification decision is made to complete the construction of the classification model of moving ships and moored ships. If the ship trajectory of a cluster in the clustering results is similar to that of a moored ship, and the AIS data shows that the ship is in a stationary or moored state, then the cluster is marked as a moored ship. If the ship trajectory of a cluster is similar to that of a moving ship, and the AIS data shows that the ship is in a moving state, then the cluster is marked as a moving ship. The navigation data processing process is as follows Figure 2 shown.

[0097] Step 2: Based on the classification model of the target ship and combined with the four-element ship model, a ship field model of dynamic and static obstacles is established.

[0098] First, a four-dimensional ship domain model is established; the four-dimensional ship domain is elliptical in shape, and the domain size is mainly determined by R fore 、R aft 、R starb 、R port Four parameters are determined, and the boundary shape of the quaternion ship domain is adjusted by these four parameters to cope with different heading scenarios. The quaternion ship domain (QSD) can be described as:

[0099] QSD={(x,y)∣f(x,y;Q)≤1,Q={R fore ,R aft ,R starb ,R port}}

[0100]

[0101] Among them, f(x, y; Q) is the function that determines the boundary of the four-dimensional ship domain, x and y are the spatial coordinates of the ship, Q is an intermediate variable, representing the four radii of the ship domain, sgn(x) and sgn(y) represent a sign function; R fore 、R aft 、R starb 、R port Respectively represent the radius of the four-element ship field in the front, back, right and left directions;

[0102] According to the established target ship classification model, the quaternary ship models of different types of target ships are adjusted, and static and dynamic quaternary ship domain models are established respectively.

[0103] If the target ship is a moving ship, it can be regarded as a dynamic ship, and the values ​​of the four radius parameters are:

[0104]

[0105] Where L is the length of the ship; k AD k is the gain coefficient of the ship's distance AD ​​in the circle, DT is the gain coefficient of the initial diameter of the cycle DT;

[0106] If the target ship is a moored ship, it can be regarded as a static obstacle and the four radius parameters can be adjusted as follows:

[0107]

[0108] Step 3: Establish local path planning based on the artificial potential field (APF) method, establish a repulsive field around the obstacle, establish a gravitational field at the target point, and introduce a global path return mechanism. Establish a gravitational field at the nodes of the global path and establish a local collision avoidance path planning algorithm that takes the return mechanism into account.

[0109] According to the artificial potential field method, a repulsive field is established around the obstacle, the field is scaled and adjusted according to the dynamic and static obstacles, a gravitational field is constructed around the target point, and a return gravitational field is established according to the path nodes.

[0110] The potential field of a ship is changing at every moment. Assume that at a certain moment, there are N s static obstacles, N d dynamic obstacles, the current sum of the artificial potential field can be expressed as the sum of the gravitational potential field and the repulsive potential field, and the formula is as follows:

[0111]

[0112] Among them, Patt is the gravitational field, P ref is the repulsive field, P g is the target point potential field, P n is the gravitational field of the global path node, P s,i is the potential field of point obstacle i, P d,j is the potential field of domain-type obstacle j.

[0113] During the local collision avoidance process, the Collision Risk Index (CRI) is used to determine when the ship begins collision avoidance and returns. When the CRI exceeds the first collision threshold, T1, the ship begins collision avoidance. By combining the current artificial potential field with map modeling, the minimum potential field near the ship can be found. The ship will tend to move toward areas with smaller potential fields, thereby achieving autonomous decision-making for local path planning and obstacle avoidance. As the ship's collision avoidance process is completed, the collision risk decreases. When the collision risk CRI falls below the second collision threshold, T2, the time to return is confirmed. At this time, the ship often experiences significant path and heading deviations. Relying solely on the target point's gravitational field makes it difficult for the ship to gradually return to the global path. Therefore, an adaptive return mechanism is established to guide the ship back to the global route.

[0114] Step 4: Develop a local collision avoidance path planning algorithm that incorporates an adaptive return-to-home mechanism. The algorithm's decisions are corrected based on the errors between each algorithm execution and planning. Adaptive gravity adjustments are implemented based on the current ship's path deviation and the degree of navigational deviation. The dynamic ship domain model and repulsive field are adjusted based on the relative motion of obstacles. The local path planning algorithm is corrected based on the errors between each planning and execution, given a given time step.

[0115] After confirming the timing of returning home, deviation detection is first performed to calculate the distance between the ship and the global path, as well as the deviation between the ship's heading and the global path heading.

[0116] The formula for the distance a ship deviates from the global path is:

[0117] Δd=||(x,y)-(x p ,y p )||

[0118] Among them, (x, y) is the current position of the ship, (x p ,y p ) is the node (nearest point) of the global path;

[0119] The formula for the angle at which a ship deviates from its global heading is:

[0120] Δψ=|ψ-ψ0|

[0121] Among them, ψ is the current heading angle of the ship, and ψ0 is the global path heading angle;

[0122] To unify the dimensions, normalization is performed, assuming that the maximum path deviation does not exceed d m , the maximum heading deviation does not exceed ψ m ;

[0123] The degree of deviation from the path is:

[0124]

[0125] The degree of deviation from the heading is:

[0126]

[0127] With the deviation parameters, the gravitational potential field can be dynamically adjusted based on the deviation error. If there is an obstacle, the dynamic ship field and repulsive potential field can be adjusted based on the relative motion state of the obstacle. An adaptive return mechanism is established. The flow chart of the adaptive return mechanism can be seen. Figure 4 , the basic steps are as follows:

[0128] Step 4.1: Initialize parameters and set targets

[0129] Assume that the speed of the ship returning after the collision is v1, the heading angle is ψ1, the path deviation is Δd1, the heading deviation is Δψ1, and the initial ship area radius is R fore , R aft , R starb , R port , the minimum encounter distance is DCPA;

[0130] To achieve the return goal, set the path deviation to be less than d min When , it can be regarded as returning to the global route, and the heading deviation is less than ψ min , it can be regarded as returning to the global heading.

[0131] Step 4.2: Adaptive Gravitational Field Adjustment

[0132] Set the time step Δt, and update the path deviation d every Δt d , heading deviation d ψ , speed deviation Δv and recalculate the weights of the return gravity field and the target gravity field. The calculation formula is as follows:

[0133]

[0134]

[0135] Among them, k1 and k2 are the attenuation degrees. The larger the value, the faster the attenuation.

[0136] The adaptive gravitational field is:

[0137] Patt =(1+ω1)P n +(1+ω2)P g

[0138] Among them, ω1 and ω2 are the weights of the return gravitational field and the target gravitational field;

[0139] During the return process, in addition to the return gravity and the target point gravity, there may also be repulsion from obstacles on the ship. Therefore, in order to better complete the return planning, it is also necessary to adjust the ship field and repulsive potential field based on the relative motion state of the obstacle and the ship.

[0140] After the ship avoids collision, the collision risk of the ship is judged based on the relative position, relative speed and relative heading of the ship. In order to better detect the collision risk, the ship area is expanded k times and an external safety distance is added;

[0141] d s =k·R

[0142]

[0143] Among them, R is the length from the center of the area to the boundary, and the k value is the multiple of the expanded range. Considering collision safety, k = 2;

[0144] According to different encounter scenarios, it is judged whether there is a collision risk between the own ship and the target ship. A radius of d is constructed with the own ship as the center. s The detection circle is adjusted according to different scenarios, such as 3 to 4 nautical miles for overtaking situations and 5 to 6 nautical miles for encounters or crossovers. A detection line is constructed based on the relative positions and headings of the ships. Whether the detection line passes through the detection circle is determined to determine whether there is a collision risk between the two ships. The degree of collision risk is determined based on the minimum distance to the closest point of approach (DCPA). The size of the dynamic ship area is then adjusted based on speed, heading, and position to enable ships to start collision avoidance in advance. The specific process is as follows:

[0145] Step 4.3: Dynamic ship field and adaptive repulsive field adjustment

[0146] Assuming R1 is the distance from the intersection of the DCPA line and the ship area to the ship, and R2 is the distance from the intersection of the DCPA extension line and the detection circle to the ship, then,

[0147]

[0148] Among them, ω DCPA Indicates the adjustment weight of the ship's area radius, is the attenuation coefficient. The larger the value, the faster the attenuation. DCPA represents the closest encounter distance between own ship and target ship.

[0149] The ship domain is adjusted based on the direction of the ship encounter. When there is a collision risk, the ship domain model is adjusted based on the direction of the target ship. The adjustment formula is as follows:

[0150] QSD e ={(x,y)|f e (x,y;Q)≤1,

[0151] Q e ={R fore +ΔR fore ,R aft +ΔR aft ,R stard +ΔR stard ,R port +ΔR port}}

[0152] ΔR r =R r ·ω DCPA

[0153] Among them, f e (x, y; Q) represents the boundary of the ship area, Q e Represents the adjustment value of the radius of the four areas, ΔR r Indicates the dynamic adjustment value of the field radius, r indicates the direction of adjustment, ΔR fore Indicates the dynamic adjustment of the forward radius of the ship, ΔR aft Indicates the dynamic adjustment of the ship's rearward radius, ΔR starb Indicates the dynamic adjustment of the starboard radius of the ship, ΔR port Indicates the dynamic adjustment of the port radius of the ship, R r ∈{R fore ,R aft ,R starb ,R port};

[0154] After avoiding collision, return to the destination. For obstacles behind, the return process needs to avoid collision with obstacles behind. Define Δv OT = The speed of the own ship divided by the speed of the obstacle behind it, expressed as:

[0155]

[0156] Among them, v OS Indicates the ship's speed, v TS Indicates the speed of the target ship;

[0157] Δθ OTis the angle between the speed directions of the own ship and the target ship, and the velocity repulsion field P of the dynamic obstacle is introduced. v , which is expressed as:

[0158]

[0159] in, is the relative velocity Δv OT The gain factor, Δv OT The smaller it is, the greater the possibility that the following ship will catch up. A larger value should be taken;

[0160] The ship's adaptive repulsive field is:

[0161]

[0162] Step 4.4: Adaptive total potential field adjustment and update

[0163] After each time step Δt, the current DCPA and relative speed Δv are recalculated. OT and the angle Δθ OT , update the current total adaptive potential field; during the ship's return process, its adaptive total potential field is:

[0164]

[0165] Among them, P att is the current total gravitational field, P ref is the current total repulsive field, P g is the target point potential field, P n is the gravitational field of the global path node, P s,i is the potential field of point obstacle i, P d,j is the potential field of domain-type obstacle j, P v is the velocity repulsive field of the dynamic obstacle, ω1 and ω2 are the update weights of the return gravitational field and the target gravitational field. Figure 3 .

[0166] Determine whether the path deviation value and heading deviation value reach the return target setting value. If not, repeat the above steps until the ship returns to the global path.

[0167] The autonomous collision avoidance path planning flow chart for the entire process of navigating from the global path, autonomously avoiding obstacles when encountering them, and returning to the global path is as follows: Figure 5 shown.

[0168] It will be easily understood by those skilled in the art that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A full-process path planning method for autonomous collision avoidance of intelligent ships, characterized in that: The following steps are involved: S1. Obtain dynamic and static information of the target ship through navigation radar and automatic identification system, establish an error distribution model of the fused target based on the dynamic and static information of the target ship, and obtain a classification model of the target ship based on the error distribution model of the fused target; S2. Based on the classification model of the target ship and the four-element ship domain model, a ship domain model of dynamic and static obstacles is established; S3. Establish local path planning based on the artificial potential field method. A repulsive field is established around obstacles in the ship domain model, and a gravitational field is established at the target point. At the same time, a global path return mechanism is introduced, and a gravitational field is established at the nodes of the global path. A local collision avoidance path planning algorithm is established that takes the return mechanism into account. S4. Adaptive gravity adjustment is implemented based on the current ship path deviation and the degree of navigation deviation. The dynamic ship domain model and repulsive field are adjusted based on the relative motion state of the obstacle. Given a time step, the local path planning algorithm is corrected according to the error of each planning and execution, forming a dynamically optimized local collision avoidance path planning algorithm that takes into account the adaptive return mechanism. The ship's path is planned based on this local collision avoidance path planning algorithm. The dynamic information of the target ship in S1 includes at least the real-time distance, speed, heading, position, longitude and latitude, and time of the target ship, and the static information includes at least the name of the target ship, the water mobile service identification code, and the navigation status.

2. The method for autonomous collision avoidance of an intelligent ship during the entire process of path planning according to claim 1 is characterized in that: S1 is specifically: Collecting raw data from each vessel using marine radar and automatic identification systems. The raw data includes at least speed, heading, longitude and latitude, and time. Generating the historical trajectory of the vessel based on the time sequence and longitude and latitude changes in the raw data. Collecting salient features of the historical trajectory data. The salient features include at least the rate of change of speed, rate of change of heading, and position change. Perform data cleaning on prominent features, check for missing values ​​and abnormal data, interpolate and fill missing values ​​based on historical trajectories, remove abnormal data, and ensure the integrity of the ship's historical trajectory data; Set the parameters of the OPTICS clustering model, including the minimum number of neighbors and the maximum neighborhood distance, and develop a metric; The normalized historical trajectory data is used as input to the OPTICS clustering model. The OPTICS clustering model is executed to automatically identify and sort the density reachability of data points, and output a variable-density clustering result. The clustering result includes at least the reachability distance and its sorting graph, as well as cluster labels. The cluster division is determined based on the reachability distance sorting graph, and the clusters are classified based on the output cluster labels. Based on the clustering results and the static information provided by the automatic identification system, the ship classification decision is made and the classification model of moving ships and anchored ships is constructed; If the ship trajectory of a cluster in the clustering results is similar to that of a moored ship, and the static information of the automatic identification system indicates that the ship is in a stationary or moored state, then the cluster is marked as a moored ship; if the ship trajectory of a cluster is similar to that of a moving ship, and its static information indicates that the ship is in a moving state, then the cluster is marked as a moving ship.

3. The method for autonomous collision avoidance path planning for an intelligent ship according to claim 1, characterized in that: The quaternary ship domain model QSD in S2 is expressed as: QSD={(x,y)|f(x,y;Q)≤1,Q={R fore ,R aft ,R starb ,R port }} Among them, f(x, y; Q) is the function that determines the boundary of the four-dimensional ship domain, x and y are the spatial coordinates of the ship, Q is an intermediate variable, representing the four radii of the ship domain; sgn(x) and sgn(y) both represent a symbolic function, which is defined as: R fore 、R aft 、R starb 、R port Represents the radius of the four-dimensional ship area in the front, back, right and left directions respectively; the four-dimensional ship area is elliptical, and its area size is represented by R fore 、R aft 、R starb 、R port Four radius parameters are determined, and the boundary shape of the quaternion ship field is adjusted by these four radius parameters to cope with different heading scenarios.

4. The method for autonomous collision avoidance path planning for an intelligent ship according to claim 3, characterized in that: The method of establishing the ship domain model for dynamic and static obstacles is as follows: according to the classification model of target ships, the four-element ship models of different types of target ships are adjusted to establish static and dynamic ship domain models respectively; If the target ship is a moving ship, it can be regarded as a dynamic obstacle, that is, a dynamic ship domain model is established for it, and the values ​​of its four radius parameters are: Where L is the length of the ship, k AD k is the gain coefficient of the ship's distance AD ​​in the circle, DT is the gain coefficient of the initial diameter of the cycle DT; If the target ship is a moored ship, it can be regarded as a static obstacle, that is, a static ship domain model is established for it, and the values ​​of its four radius parameters are:

5. The method for full-process path planning of autonomous collision avoidance for intelligent ships according to claim 4 is characterized in that: S3 specifically: According to the artificial potential field method, a repulsive field is established around the obstacle, the field is scaled and adjusted according to the radius parameters of dynamic and static obstacles, a gravitational field is constructed around the target point, and a return gravitational field is established according to the path nodes; The potential field of a ship is changing at every moment. Assume that at a certain moment, there are N s static obstacles, N d dynamic obstacles, the current sum of the artificial potential field is the sum of the gravitational potential field and the repulsive potential field, expressed as: Among them, P f is the sum of the current artificial potential field, P att is the gravitational field, P ref is the repulsive field, P g is the target point potential field, P n is the gravitational field of the global path node, P s，i is the potential field of point obstacle i, P d，j is the potential field of domain-type obstacle j; During the local collision avoidance process, the collision risk index (CRI) is used to determine the time points for the ship to start collision avoidance and return. When the CRI is higher than the first collision threshold, the ship begins collision avoidance. By combining the current artificial potential field sum with map modeling, the minimum potential field near the ship is found. The ship will tend to move towards the area with smaller potential field. As the collision avoidance process is completed, the collision risk decreases. When the CRI is lower than the second collision threshold, the time to return is confirmed, thus realizing local path planning and autonomous decision-making for obstacle avoidance.

6. The method for full-process path planning of autonomous collision avoidance for intelligent ships according to claim 5, characterized in that: S4 is specifically: After confirming the timing of returning home, deviation detection is performed to calculate the distance between the ship and the global path, as well as the deviation between the ship's heading and the global path heading; The formula for the distance that a ship deviates from the global path is expressed as: Δd=||(x,y)-(x p ,y p )|| Among them, Δd is the distance from the ship to the global path, (x, y) is the current position of the ship, (x p ,y p ) is the global path node closest to the ship; The formula for the angle at which a ship deviates from its global heading is: Δψ=|ψ-ψ0| Where Δψ is the deviation between the ship heading and the global path heading, ψ is the current heading angle of the ship, and ψ0 is the global path heading angle; To unify the dimensions, normalization is performed, assuming that the maximum path deviation does not exceed d m , the maximum heading deviation does not exceed ψ m ; The degree of deviation from the path d d Expressed as: Degree of deviation from the course d ψ Expressed as: Dynamically adjust the gravitational potential field based on the deviation error. If there is an obstacle, adjust the dynamic ship field and repulsive potential field based on the relative motion state of the obstacle. Establish an adaptive return mechanism, including: parameter and target initialization, adaptive gravitational field adjustment, dynamic ship field and adaptive repulsive field adjustment, adaptive total potential field adjustment and update.

7. The method for full-process path planning of autonomous collision avoidance for intelligent ships according to claim 6, characterized in that: The parameters and target initialization are as follows: Assume that the speed at the moment of returning after the collision is v1, the heading angle is ψ1, the path deviation is Δd1, the heading deviation is Δψ1, and the initial ship area radius is R fore , R aft , R starb , R port , the minimum encounter distance is DCPA; To achieve the return goal, set the path deviation to be less than d min When , it is considered to return to the global route; Course deviation is less than ψ min When , it is regarded as returning to the global heading; The specific steps of adaptive gravitational field adjustment are: Set the time step Δt, and update the path deviation d every Δt d , heading deviation d ψ , speed deviation Δv and recalculate the weights of the return gravity field and the target gravity field. The calculation formula is expressed as: Among them, k1 and k2 are the attenuation degrees. The larger the value, the faster the attenuation. The adaptive gravitational field is expressed as: P att =(1+ω1)P n +(1+ω2)P g Among them, ω1 and ω2 are the weights of the return gravitational field and the target gravitational field.

8. The method for full-process path planning of autonomous collision avoidance for intelligent ships according to claim 7, characterized in that: The specific steps for dynamic ship field and adaptive repulsive field adjustment are: After the ship avoids collision, the collision risk of the ship is judged according to the relative position, relative speed and relative heading of the ship. In order to better detect the collision risk, the ship area is expanded k times and an external safety distance d is added. s , expressed as: d s =k·R in, Indicates the boundary detection range of the expanded ship area, R is the length from the center of the area to the boundary, and k is the multiple of the expanded range; According to different encounter scenarios, it is judged whether there is a collision risk between the own ship and the target ship. A radius of d is constructed with the own ship as the center. s The detection circle is adjusted according to the different scenes; A detection line is constructed based on the relative positions and headings of the ships. Whether the detection line passes through the range of the detection circle is used to determine whether there is a collision risk between the two ships. The closest approach distance (DCPA) is one of the main factors in determining the risk of collision. It represents the distance between the two ships when they are in their current state of motion and continue to sail at a predetermined heading and speed. The smaller the DCPA value, the greater the collision risk. The size of the dynamic ship area is adjusted based on speed, heading, and position to enable ships to start collision avoidance in advance. The specific steps are as follows: Assuming R1 is the distance from the intersection of the DCPA line and the ship area to the ship, and R2 is the distance from the intersection of the DCPA extension line and the detection circle to the ship, then, Among them, ω DCPA Indicates the adjustment weight of the ship's field radius; is the attenuation coefficient. The larger its value, the faster the attenuation. DCPA represents the closest approach distance between own ship and target ship. Adjust the ship domain based on the direction of the ship encounter; when there is a collision risk, adjust the ship domain model QSD based on the direction of the target ship e , the adjustment formula is expressed as: QSD e ={(x,y)|f e (x,y;Q)≤1, Q e ={R fore +ΔR fore ,R aft +ΔR aft ,R starb +ΔR starb ,R port +ΔR port }} ΔR r =R r ·oh DCPA Among them, f e (x, y; Q) represents the boundary of the ship area, Q e Represents the adjustment value of the radius of the four areas, ΔR r Indicates the dynamic adjustment value of the field radius, r indicates the direction of adjustment, ΔR fore Indicates the dynamic adjustment of the forward radius of the ship, ΔR aft Indicates the dynamic adjustment of the ship's rearward radius, ΔR starb Indicates the dynamic adjustment of the starboard radius of the ship, ΔR port Indicates the dynamic adjustment of the port radius of the ship, R r ∈{R fore , R aft , R starb , R port }; After avoiding collision, return to the destination. For obstacles behind, the return process needs to avoid collision with obstacles behind. Define Δv OT = is the speed of own ship divided by the speed of the obstacle behind, that is Among them, v OS Indicates the ship's speed, v TS Indicates the speed of the target ship; Δθ OT is the angle between the speed directions of the own ship and the target ship, and the velocity repulsion field P of the dynamic obstacle is introduced. v , which is expressed as; in, is the relative velocity Δv OT The gain factor, Δv OT The smaller it is, the greater the possibility of the obstacle behind catching up. A larger value should be taken; Adaptive repulsive field P of the ship ref for:

9. The method for full-process path planning of autonomous collision avoidance for intelligent ships according to claim 8, characterized in that: The specific steps of adaptive total potential field adjustment and update are: After each time step Δt, the current DCPA and relative speed Δv are recalculated. OT and the angle Δθ OT , update the current total adaptive potential field; during the ship's return process, its adaptive total potential field is: Among them, P att is the current total gravitational field, P ref is the current total repulsive field, P g is the target point potential field, P n is the gravitational field of the global path node, P s，i is the potential field of point obstacle i, P d，j is the potential field of domain-type obstacle j, P v is the velocity repulsion field of the dynamic obstacle, ω1 and ω2 are the update weights of the return gravitational field and the target gravitational field.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 9 are implemented.

Images