A data-driven based global path planning method for a ship

Abstract

The application relates to a data-driven global path planning method for a ship, which comprises the following steps: obtaining ship trajectory data based on preprocessed AIS data, constructing a trajectory space-time sequence feature database based on the ship trajectory data, constructing a grid environment model of a target water area based on the trajectory space-time sequence feature database, loading parameter information of a target ship and position information of a target point into the grid environment model for grid environment matching, and determining a global optimal path of the target ship to the target point in the grid environment model based on an improved A* algorithm. The application applies water area traffic features to the planning of a global path of a ship, and improves the safety of ship navigation.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of ship path planning, and particularly relates to a ship global path planning method based on data driving. BACKGROUND

[0002] With the continuous development of water transportation, intelligent shipping is an inevitable trend of future development of water transportation, and is also the top priority of the development and research of China's shipping industry. With the gradual recovery of shipping demand and the application of autonomous ships, the navigation path and environment of ships will inevitably be more complex and diverse. Therefore, how to plan a navigation path for a ship in a complex marine environment that can achieve or approach the optimal route and avoid factors affecting navigation safety, to meet the navigation needs of the ship, has become increasingly important.

[0003] Ship global path planning mainly considers the characteristics of static obstacles such as navigation environment, and plans a ship navigation path from the starting point to the termination point, which includes turning points and heading and other elements. In view of the development needs of ship navigation safety and efficiency and intelligent ships, researching the global path planning method of ships is of great significance to improve the intelligent level and operational efficiency of ships.

[0004] However, the current ship global path planning algorithm only plans the path from the water environment, without considering the water ship traffic characteristics, which may result in a global path that crosses the water route frequently, increasing the risk of water ship traffic. SUMMARY

[0005] Therefore, it is necessary to provide a ship global path planning method based on data driving to solve the problem of not considering water ship traffic characteristics in the prior art global path planning.

[0006] The present application provides a ship global path planning method based on data driving, comprising:

[0007] Obtaining ship trajectory data based on preprocessed AIS data, and constructing a trajectory spatio-temporal sequence feature database based on the ship trajectory data;

[0008] Constructing a grid environment model of the target water area based on the trajectory spatio-temporal sequence feature database;

[0009] Loading the obtained parameter information of the target ship and the position information of the target point into the grid environment model for grid environment matching;

[0010] Determining the global optimal path of the target ship to the target point in the grid environment model based on the improved A* algorithm.

[0011] In some possible implementation manners, the AIS data after preprocessing is used to obtain ship trajectory data, a trajectory spatio-temporal sequence feature database is constructed based on the ship trajectory data, and the construction includes:

[0012] The noise data in the historical AIS data of the target water area ship is deleted or corrected to obtain the AIS data after preprocessing;

[0013] The trajectory data set of the target water area ship is formed by sorting the AIS data after preprocessing according to time sequence;

[0014] The trajectory spatio-temporal sequence feature database is established based on the trajectory data set of the target water area ship.

[0015] In some possible implementation manners, a grid environment model of the target water area is constructed based on the trajectory spatio-temporal sequence feature database, and the construction includes:

[0016] The chart information of the target water area is extracted by using a pixel value extraction method, and a grid map is established;

[0017] The historical trajectory data in the trajectory spatio-temporal sequence feature database is subjected to trajectory similarity measurement and a trajectory clustering algorithm to obtain different historical trajectory clusters of the target water area ship route;

[0018] The historical trajectory clusters are normalized and projected into the grid map respectively.

[0019] In some possible implementation manners, the global optimal path of the target ship to the target point in the grid environment model is determined based on the improved A* algorithm, and the determination includes:

[0020] Step 1, the node of the target ship in the grid environment model is set as a starting point and added into an Open table, and the obstacle points in the grid environment model are put into a Closed table;

[0021] Step 2, whether the extended node in the Open table has a target point is checked based on an extended neighborhood node, if not, the step 3 is jumped to, if yes, the target point is added into the Closed table, the search path is saved, and the algorithm is ended;

[0022] Step 3, the current node is set as a parent node, and is transferred to the Closed table;

[0023] Step 4, the value of each extended node is calculated according to an evaluation function of the target ship, and the smallest point is selected as a new node, and the remaining extended nodes are stored into the Closed table;

[0024] Step five, confirming whether the Open table is empty, if not, return to step two, if the Closed table is empty, the path does not exist, and the algorithm ends.

[0025] In some possible implementation manners, the method further includes: determining the extended neighborhood node through the movable direction of the target ship and the number of neighborhoods;

[0026] The formula for calculating the movable direction of the target ship and the number of neighborhoods is as follows:

[0027]

[0028]

[0029] In the formula, K is the number of layers of the node, D K is the movable direction of the target ship, N K is the number of neighborhoods, and Z represents an integer.

[0030] In some possible implementation manners, the formula of the evaluation function of the target ship is f(n) = g(n) + h(n).

[0031] In the formula, g(n) is a cost function with a reward and punishment factor, and h(n) is a heuristic function with a reward and punishment factor.

[0032] In some possible implementation manners, the reward and punishment factor includes a first penalty factor, a reward factor and a second penalty factor, and the function expression corresponding to the reward and punishment factor is:

[0033]

[0034] In the formula, U all is the function value of the reward and punishment factor, U att is the function value of the reward factor of the target point to the target ship, U rep is the function value of the first penalty factor of the obstacle to the target ship, U other represents the function value of the second penalty factor of the target ship to other factors except the obstacle, k1 + k2 + k3 = 1, and k1, k2 and k3 are the weights of the function of the reward and punishment factor, the function of the first penalty factor and the function of the second penalty factor respectively.

[0035] In some possible implementation manners, the method further includes: determining the function of the first penalty factor according to the penalty value of the obstacle to the target ship, and determining the function of the second penalty factor according to the penalty value of other factors except the obstacle to the target ship.

[0036] The function expression of the first penalty factor is:

[0037]

[0038] wherein: U rep (X) is a penalty value generated by the obstacle to the target ship, k rep represents a gain coefficient, X0represents position coordinate information of the obstacle, X c represents position coordinate information of the target ship, d0represents an influence range of the obstacle, d(X c , X0) represents a distance value from the target ship to the obstacle;

[0039] The function expression of the second penalty factor is:

[0040] U other = k * p(x, y)

[0041] wherein: U other represents a penalty value of the target ship affected by other factors except the obstacle, p(x, y) represents a probability of the target ship affected by other factors except the obstacle to have a safety accident, and k represents a gain coefficient.

[0042] In some possible implementation manners, the value of the function of the reward factor is equal to the inverse of the density value of the historical trajectory cluster after normalization processing.

[0043] In some possible implementation manners, the cost function is determined according to actual path cost from a starting point to the position of the target ship, and the heuristic function is determined according to estimated path cost from the position of the target ship to the position of the target point.

[0044] The heuristic function is determined according to estimated path cost from the position of the target ship to the position of the target point.

[0045] The calculation formula of the heuristic function is:

[0046]

[0047] wherein: C is path cost of horizontal or lateral movement of one grid, min(U all ) is a minimum reward and penalty value, x n , y n and x goal , y goal are position coordinate information of the target ship and position coordinate information of the target point respectively.

[0048] The calculation formula of the g(n) function is:

[0049]

[0050] wherein: C is path cost of horizontal or lateral movement of one grid, min(Uall is the minimum reward and punishment value, x n , y n and x goal , y goal are position coordinate information of the target ship and position coordinate information of the target point, respectively.

[0051] The beneficial effects of the above embodiment are that the ship global path planning method based on data driving provided by the application firstly pre-processes historical AIS data, and constructs a trajectory behavior feature database based on time sequence; then, based on an electronic chart, a pixel value extraction method is used to construct a basic grid environment model, and combined with grid ship traffic flow data, a water area grid environment model based on trajectory correlation features is established. Finally, based on the water area grid environment model, the A* algorithm is improved by combining ship behavior features and navigation specifications, so that the safety of navigation is improved in ship global path planning. BRIEF DESCRIPTION OF DRAWINGS

[0052] Figure 1 is a flowchart of one embodiment of the ship global path planning method based on data driving provided by the application;

[0053] Figure 2 is a flowchart of one embodiment of step S101 in Figure 1 ;

[0054] Figure 3 is a flowchart of one embodiment of step S102 in Figure 1 . DETAILED DESCRIPTION

[0055] The preferred embodiments of the application will be described in detail below with reference to the accompanying drawings, wherein the drawings constitute a part of this application and serve to explain the principles of the embodiments of the application, but are not used to limit the scope of the application.

[0056] Figure 1 is a flowchart of one embodiment of the ship global path planning method based on data driving provided by the application, as shown in Figure 1 , a ship global path planning method based on data driving comprises the following steps:

[0057] S101, obtaining ship trajectory data based on pre-processed AIS data, and constructing a trajectory space-time sequence feature database based on the ship trajectory data;

[0058] S102, constructing a grid environment model of a target water area based on the trajectory space-time sequence feature database;

[0059] S103, load the acquired parameter information of the target ship and the position information of the target point to the grid environment model for grid environment matching;

[0060] S104, determine the global optimal path of the target ship to the target point in the grid environment model based on the improved A* algorithm.

[0061] Compared with the prior art, the ship global path planning method based on data driving provided by the embodiment firstly pre-processes historical AIS data, and constructs a trajectory behavior feature database based on time sequence; then, based on an electronic chart, a pixel value extraction method is used to construct a basic grid environment model, and the grid ship traffic flow data is combined for normalized modeling, and a water area grid environment model based on trajectory correlation features is established. Finally, based on the water area grid environment model, the A* algorithm is improved in combination with ship behavior features and navigation specifications, so that the safety of navigation is improved in ship global path planning.

[0062] In step S103, the parameter information of the target ship includes the size, speed, heading, position information of the target ship, and other related information such as gain coefficient, algorithm step length and maximum turning angle of the ship.

[0063] In some embodiments of the present application, in step S101, the ship trajectory data is obtained based on the pre-processed AIS data, and a trajectory space-time sequence feature database is constructed based on the ship trajectory data, including:

[0064] S201, deleting or correcting noise data in the historical AIS data of the target water area ship to obtain pre-processed AIS data;

[0065] S202, the trajectory data set of the target water area ship formed by sorting the pre-processed AIS data according to time sequence;

[0066] S203, establishing a trajectory space-time sequence feature database based on the trajectory data set of the target water area ship.

[0067] In step S201, the pre-processed AIS data includes deleting abnormal water mobile communication service identification code, completely same repeated data and data noise in the historical AIS data of the target water area ship, and supplementing, correcting or deleting abnormal latitude and longitude, abnormal speed and heading data, to obtain the pre-processed AIS data. The pre-processed AIS data is:

[0068] Static(i)=(MMSI,callsign,IMO,type,len,bre,dra)(i=1,2,...,n)

[0069] Activity(i) = (time, MMSI, lon, lat, sog, cog,...) (i = 1, 2,..., N)

[0070] wherein Static(i) is the i-th ship static information data, MMSI is the ship's mobile service identity, callsign, IMO, type, len, bre, dra represent the call sign, IMO number, type, length, width, draft of the ship to which the track point belongs respectively, Activity(i) is the i-th ship dynamic information data, MMSI is the ship's mobile service identity, time, lon, lat, len, sog, cog represent the behavior characteristics and space-time information of the ship at the point respectively

[0071] …

[0072] such as time, MMSI, latitude and longitude, ground speed and ground course.

[0073] In step S202, the continuous AIS data information of a single ship in a certain time period is obtained, i.e., a ship time sequence track, and the ship time sequence track is combined to obtain a track data set of ships in a target water area in a certain time period.

[0074] The track data set includes extraction of track feature points and classification of sub-track segments based on time, speed, heading, and ship latitude and longitude information. First, feature points of track segments are extracted based on time and distance, and track segments of different ships with excessively large space-time distance are divided into several sub-track segments. Then, behavior characteristic parameters of each sub-track segment are extracted based on speed, heading, and relative position, and the sub-track segments are further divided into stay feature track segments and navigation feature track segments.

[0075] In the specific implementation of the present application, the size of the grid is selected as one time the length of the target ship, so that the position of the navigation hazard can be accurately represented, and the grid division and model establishment are performed in this way; the obtained historical track information is introduced to guide the ship to avoid the low-frequency navigation area of the ship and reduce the collision risk with the navigation hazard.

[0076] In step S203, after the data in the track data set is processed by de-duplication, abnormality screening, etc., an AIS static information data table and an AIS dynamic information data table are constructed, and an association between the two data tables is established with MMSI as the index. Further, semantic matching is performed on the ship type in the AIS static information, and further segmentation is performed on the behavior characteristic data in the AIS dynamic information, and finally the processed data is stored in the AIS static information data table and the AIS dynamic information data table, to obtain a track space-time sequence feature database.

[0077] In some embodiments of the present application, in step S102, the trajectory spatio-temporal sequence feature database is used to construct a grid environment model of the target water area, comprising:

[0078] S301, using a pixel value extraction method, extracting chart information of the target water area and establishing a grid map;

[0079] S302, performing trajectory similarity measurement and trajectory clustering algorithm on the historical trajectory data in the trajectory spatio-temporal sequence feature database to obtain different historical trajectory clusters of the target water area ship route;

[0080] S303, projecting the historical trajectory clusters after normalization processing into the grid map respectively.

[0081] It should be noted that the chart provides very important information support in ship route planning and ship navigation decision-making. The chart can provide all information including land, shoal, rock, wreck and bridge and other navigation-obstructing objects, and wind, wave, current, water depth and other factors threatening the safety of ship navigation in ship navigation activities. Different types of navigation information are color-coded in the chart, so different color classifications can be used to extract and identify navigation-obstructing objects and water depth information in the chart.

[0082] In step S301, after obtaining the processed chart matrix, on the basis of the existing grid modeling method, the grid corresponding latitude and longitude coordinates are matched with the latitude and longitude corresponding RGB values in the chart image, and the RGB values of the chart pixels are mapped with the corresponding chart information type. Different types of pixel points are assigned values and used as grid information to establish a grid map.

[0083] It should be noted that Dynamic Time Warping (DTW) is a widely used speech recognition technology. The basic idea is to twist and adjust the time of two trajectory sequences, locally scale and align repeated points on the time axis, etc. to make the two trajectory shapes as consistent as possible, and then find the minimum transformation path between the two trajectories by dynamic programming. Since the DTW algorithm has less length limit on the trajectory, it meets the characteristics of discrete sampling of ship trajectories, and can distinguish trajectories in different directions. Therefore, in step S302, the DTW algorithm is used to measure the trajectory similarity of the historical trajectory data. Specifically, for two trajectories Traj1 and Traj2, the dynamic time warping calculation formula is as follows:

[0084]

[0085] In the formula: DTW (Traj1, Traj2) is the DTW distance between all trajectory points of the two trajectories Traj1 and Traj2, Head (Traj) represents the first point of the trajectory, and rest (Traj) represents all trajectory points of the trajectory except the first point.

[0086] It should be noted that the hierarchical clustering algorithm can be divided into two categories according to its clustering order, from bottom to top and from top to bottom. The former is to start from the bottom layer, calculate the distance between the clustering targets, and gradually merge the most similar clustering targets to form a new layer of data, then calculate the distance between each cluster in the newly formed layer, and iterate until all data is clustered into one class or the termination condition is reached. The latter is the opposite of the former, which sets all clustering targets as the same cluster, then divides the initial cluster into sub-clusters, and continues to divide the sub-clusters until the single-node cluster is reached. The hierarchical clustering algorithm has strong interpretability, but the time complexity of the algorithm is high. Therefore, in step S302, the hierarchical clustering algorithm in the trajectory clustering algorithm is selected to cluster the historical trajectory data.

[0087] In order to guide the target ship to avoid the low-frequency navigation area of the ship and reduce the risk of collision with the navigation obstacle, in step S303, the grid of the historical trajectory data is normalized, and the habitual route of the ship can be obtained by clustering the historical trajectory data, which can provide certain reference and reference for the planning of the ship path. Therefore, the historical ship route is regarded as an attractive area. At the same time, the trajectory point density of the ship is normalized, and the normalized grid density value is assigned to the grid, so as to conform to the grid environment information. The normalization formula of the trajectory point density is as follows:

[0088]

[0089] In the formula: x is the density of the point, x max is the maximum value of the density.

[0090] It should be noted that the A*(A-Star) algorithm is a most effective direct search method for solving the shortest path in a static road network, and is also an effective algorithm for solving many search problems. The formula of the A*(A-Star) algorithm is f(n) = g(n) + h(n), wherein f(n) is the minimum cost estimate from the initial state to the target state via the state n, g(n) represents the minimum cost from the initial state to the state n, and h(n) represents the minimum estimated cost of the path from the state n to the target state. Two state tables are used in the A*(A-Star) algorithm, which are referred to as the Open table and the Closed table. The Open table is composed of nodes to be examined, and the Closed table is composed of nodes that have been examined. The A*(A-Star) algorithm used in the present embodiment uses f(n) to represent the path cost from the starting point to the target point via the current point n, g(n) to represent the actual path cost from the starting point to the current point n, and h(n) to represent the estimated cost from the current point n to the target point.

[0091] In some embodiments of the present application, the improved A* algorithm determines the globally optimal path of the target ship to the target point in the grid environment model, comprising:

[0092] Step one, setting the node of the target ship in the grid environment model as a starting point and adding it into the Open table, and putting the obstacle points in the grid environment model into the Closed table;

[0093] Step two, checking whether the expanded node in the Open table has the target point based on the expanded neighborhood node, if not, jumping to step three, if yes, adding the target point into the Closed table, saving the search path and ending the algorithm;

[0094] Step three, setting the current node as a parent node and transferring it to the Closed table;

[0095] Step four, calculating the value of each expanded node according to the evaluation function of the target ship, and selecting the smallest point as a new node, while storing the remaining expanded nodes into the Closed table;

[0096] Step five, checking whether the Open table is empty, if not, returning to step two, if yes, the path does not exist, and the algorithm ends.

[0097] In order to effectively improve the feasibility of the planned global path, reduce the length and number of nodes in the global path, and make the planned global path tend to be smooth, the neighborhood nodes in the A*(A-Star) algorithm are expanded. In some embodiments of the present application, the expanded neighborhood nodes are determined by the movable direction of the target ship and the number of neighborhoods.

[0098] Wherein, the target ship movable direction and the neighborhood quantity calculation formula are as follows:

[0099]

[0100]

[0101] In the formula, K is the number of layers of nodes, D K The target ship movable direction, N K The neighborhood quantity, Z represents an integer.

[0102] If the distance between the ship and the bank is too close during the movement of the ship, the bank wall effect will occur, the bank suction and bank pushing phenomenon will occur, the maneuverability of the ship is reduced, and the probability of accidents is increased, which is not conducive to the navigation safety of the ship. The collision risk between the ship and the obstacle should be considered during the navigation of the ship, and a safe distance should be kept from the obstacle, so a penalty factor function is designed. In some embodiments of the present application, the formula of the evaluation function of the target ship is: f(n)=g(n)+h(n); wherein, g(n) is a cost function with a reward and punishment factor, and h(n) is a heuristic function with a reward and punishment factor.

[0103] The navigation safety of the ship is also affected by weather conditions such as wind, wave, current and tide, etc. For example, when there is a typhoon in the navigation water area, the ship should avoid entering the typhoon influence range to ensure the navigation safety of the ship; when there is a poor visibility weather such as heavy fog in the port water area, the ship should be cautious driving or even be stranded. For the above influences, a second penalty factor function is designed to represent them. In some embodiments of the present application, the reward and punishment factor includes a first penalty factor, a reward factor and a second penalty factor, and the function expression corresponding to the reward and punishment factor is:

[0104]

[0105] In the formula: U all The function value of the reward and punishment factor, U att The function value of the target point to the target ship reward factor, U rep The function value of the first penalty factor generated by the obstacle to the target ship, U other The function value of the second penalty factor of the target ship affected by other factors except the obstacle, k1+k2+k3=1, k1, k2, k3 are the weights of the function of the reward and punishment factor, the function of the first penalty factor and the function of the second penalty factor.

[0106] In some embodiments of the present application, the first penalty factor is determined according to a function of a penalty value generated by the obstacle to the target ship, and the second penalty factor is determined according to a function of a penalty value generated by other factors to the target ship;

[0107] The function expression of the first penalty factor is:

[0108]

[0109] In the formula, U rep (X) is a penalty value generated by the obstacle to the target ship, k rep represents a gain coefficient, X0 represents position coordinate information of the obstacle, X c represents position coordinate information of the target ship, d0 is an influence range of the obstacle, d(X c , X0) represents a distance value from the target ship to the obstacle;

[0110] The function expression of the second penalty factor is:

[0111] U other = k * p(x, y)

[0112] In the formula, U other represents a penalty value of the target ship to other factors, p(x, y) represents a probability of a safety accident of the target ship affected by other factors, and k represents a gain coefficient.

[0113] In some embodiments of the present application, the value of the function of the reward factor is equal to the inverse of the density value of the historical trajectory cluster after normalization processing.

[0114] In order to ensure the safety distance between the target ship and the obstacle, so as to ensure the navigation safety of the target ship, it is necessary to introduce the reward and punishment function into the actual cost of ship movement, that is, to add it in the g(n) function. In order to improve the accuracy of the heuristic function h(n) in estimating the actual moving cost and ensure the efficiency of the algorithm, the reward and punishment factor function also needs to be included in the estimated cost in the heuristic function h(n). In some embodiments of the present application, the cost function determined by the actual path cost from the starting point to the position of the target ship and the heuristic function determined by the estimated path cost from the position of the target ship to the target point are further included.

[0115] The calculation formula of the heuristic function is:

[0116]

[0117] In the formula, C is the cost of moving one grid horizontally or laterally, min(U all) is the minimum reward and punishment value, x n , y n and x goal , y goal are position coordinate information of the target ship and position coordinate information of the target point respectively.

[0118] The cost function calculation formula is:

[0119]

[0120] In the formula, C is the path cost of moving one grid horizontally or laterally, min(U all ) is the minimum reward and punishment value, x n , y n and x goal , y goal are position coordinate information of the target ship and position coordinate information of the target point respectively.

[0121] Those skilled in the art can understand that all or part of the processes of the above-mentioned embodiments can be completed by a computer program instructing relevant hardware, and the program can be stored in a computer readable storage medium. The computer readable storage medium is a disk, an optical disk, a read-only memory or a random access memory, etc.

[0122] The above is only a preferred specific embodiment of the present application, but the protection scope of the present application is not limited to this. Any person skilled in the art can easily think of changes or replacements within the technical range disclosed by the present application, which should be covered in the protection scope of the present application.

Claims

1. A data-driven based global path planning method for a ship, characterized in that, The application relates to a ship trajectory planning method based on a grid environment model. The application comprises the following steps: Based on preprocessed AIS data, ship trajectory data is obtained, and a trajectory space-time sequence feature database is constructed based on the ship trajectory data; A grid environment model of a target water area is constructed based on the trajectory space-time sequence feature database; Based on the improved A An algorithm determines a globally optimal path for the target ship to reach the target point in the grid environment model; The improved A The algorithm determines a globally optimal path for the target ship to reach the target point in the grid environment model, comprising: Parameter information and position information of a target ship are loaded into the grid environment model for grid environment matching; Step 1: a node of the target ship in the grid environment model is set as a starting point and is added into an Open table, and obstacle points in the grid environment model are put into a Closed table; Step 2: based on an extended neighbor node, it is checked whether an extended node in the Open table has a target point, if not, step 3 is jumped to, if yes, the target point is added into the Closed table, a search path is saved, and the algorithm is ended; Step 3: a current node is set as a parent node, and is transferred to the Closed table; Step 4: a value of each extended node is calculated according to an evaluation function of the target ship, and a smallest point is selected as a new node, and the rest of the extended nodes are put into the Closed table; The formula of the evaluation function of the target ship is: ; wherein, g ( n ) is a cost function with reward and penalty factors, h ( n ) is a heuristic function with reward and penalty factors; Step 5: whether the Open table is empty is confirmed, if not, step 2 is returned, if yes, the path does not exist, and the algorithm is ended; In the formula: is a function value of the reward factor, is a function value of the reward factor generated by the target point to the target ship, is a function value of the first penalty factor generated by the obstacle to the target ship, is a function value of the second penalty factor of the target ship by other factors except the obstacle, , are weights of the functions of the reward factor, the first penalty factor and the second penalty factor, respectively. The reward and punishment factors include a first punishment factor, a reward factor and a second punishment factor, and a function expression corresponding to the reward and punishment factors is: The application further comprises a function of the first punishment factor determined according to a punishment value of the target ship generated by an obstacle, and a function of the second punishment factor determined according to a punishment value of the target ship generated by other factors except the obstacle; In the formula: is a penalty value generated by the obstacle to the target ship, represents is a corresponding gain coefficient, represents position coordinate information of the obstacle, is position coordinate information of the target ship, is an influence range of the obstacle, represents a distance value of the target ship to the obstacle; The function expression of the first punishment factor is: wherein: represents a penalty value for the target ship affected by other factors than obstacles, represents a probability of a safety incident for the target ship affected by other factors than obstacles, k represents a gain coefficient; The function expression of the second punishment factor is: The application further comprises a cost function determined according to an actual path cost from a starting point to the position of the target ship, and a heuristic function determined according to an estimated path cost from the position of the target ship to the position of a target point; The calculation formula of the heuristic function is: wherein: C is the cost of a path moving horizontally or laterally one grid, is the minimum penalty value, and are position coordinate information of the target ship and position coordinate information of the target point, respectively.

2. The data-driven based global path planning method for a ship according to claim 1, wherein, The calculation formula of the cost function is: The application obtains ship trajectory data based on preprocessed AIS data, and constructs a trajectory space-time sequence feature database based on the ship trajectory data, and the method comprises the following steps: Noise data in historical AIS data of a target water area ship is deleted or corrected to obtain preprocessed AIS data; The preprocessed AIS data is sorted according to a time sequence to form a trajectory data set of the target water area ship; 3. The data-driven based global path planning method for a ship according to claim 1, wherein, A trajectory space-time sequence feature database is established based on the trajectory data set of the target water area ship. The application constructs a grid environment model of a target water area based on a trajectory space-time sequence feature database, and the method comprises the following steps: A pixel value extraction method is used to extract chart information of the target water area and establish a grid map; Trajectory similarity measurement and a trajectory clustering algorithm are performed on historical trajectory data in the trajectory space-time sequence feature database to obtain different historical trajectory clusters of a target water area ship route; 4. The data-driven based global path planning method for a ship according to claim 1, wherein, The historical trajectory clusters are normalized and projected into the grid map respectively. The application further comprises the following steps: The extended neighbor node is determined by a movable direction of the target ship and a number of neighbors. The target ship movable direction and the calculation formula of the neighborhood quantity are as follows: In the formula, K is the number of layers of nodes, is the movable direction of the target ship, is the number of neighborhoods, and Z represents an integer.

5. The data-driven based global path planning method for a ship according to claim 3, wherein, The value of the function of the reward factor is equal to the inverse of the density value of the historical trajectory cluster after normalization.

Images