An AIS route segment clustering based route planning method

Abstract

The application provides a route planning method based on AIS route segment clustering, comprising: obtaining ship AIS data and ship specification data; performing clustering analysis on the ship specification data according to navigable factors; extracting ship connecting route segments between two ports based on AIS data of different navigable types; constructing a ship route segment connectivity graph for sailing between two ports and feature information of the route segments based on the obtained ship route segments, and obtaining an optimal route by combining distance, arrival time, heat, and energy consumption information. The application classifies ships by clustering analysis on ship navigable factors to obtain ship navigable type clusters, extracts connecting route segments between two ports for different type ships, and proposes a route planning algorithm based on route segments, thereby retaining the continuity between route track points and points, improving ship sailing efficiency, and ensuring sailing safety.

Description

Technical Field

[0001] This invention relates to the field of route planning technology, and more particularly to a route planning method based on AIS route segment clustering. Background Technology

[0002] Route planning aims to provide vessels with the optimal route from origin to destination while meeting design specifications and navigational safety requirements. The accuracy of route planning has a significant impact on vessel safety. Because vessels in different waters have different characteristics based on their shipping, trade, and service requirements, such as ship type, tonnage, and navigation direction, they typically have different commonly used routes. The development of maritime trade has led to increasingly busy shipping traffic and more complex navigation waters. Complex sea conditions have a significant impact on navigational safety and pose challenges to route planning. Therefore, building models using environmental characteristic data is an effective method to improve the accuracy of route planning. To ensure navigational safety, analyzing track characteristic information using AIS data and vessel specification data, identifying current navigational environment patterns based on sea state data, and intelligently planning target routes to ensure navigational safety has become an important task.

[0003] Automatic Identification System (AIS) is one of the most important systems in intelligent shipping systems, covering almost all vessels at sea. AIS can send and receive AIS information, providing dynamic and static data on vessels, including position, speed, and heading, supporting the Vessel Traffic Service (VTS) system. Simultaneously, AIS can receive AIS data from other vessels within 20 nautical miles, thus understanding the positions of surrounding vessels. AIS data not only helps improve vessel safety and efficiency but can also be used for maritime traffic surveys and analysis. By collecting and analyzing AIS data, the navigation characteristics of different types of vessels at sea can be explored, navigation methods in specific areas and the experience of route planners can be discovered, thereby revealing general patterns and rules of vessel navigation. AIS data is an important data source for studying maritime traffic and a crucial basis for optimizing maritime traffic. Therefore, the comprehensive navigation environment information and traffic control management information provided by AIS data better reflects the true navigation status of vessels at sea.

[0004] Currently, route planning schemes based on AIS data mainly rely on extracting turning points from AIS data to form a connected graph. This method ignores the continuity between points, resulting in low accuracy in route planning. Summary of the Invention

[0005] To address the technical problem that the proposed method of extracting turning points from AIS data to form a connected graph ignores the continuity between points, this invention provides a route planning method based on AIS route segment clustering. This invention classifies ships by obtaining navigable type clusters through cluster analysis of ship navigability factors, extracts connected route segments between two ports for different types of ships, and proposes a route planning algorithm based on route segments, thereby preserving the continuity between points on the route trajectory.

[0006] The technical means employed in this invention are as follows:

[0007] A route planning method based on AIS route segment clustering includes:

[0008] Acquire vessel AIS data and vessel specification data; perform cluster analysis on vessel specification data based on navigability factors; extract vessel connectivity routes between two ports based on AIS data of different navigability types;

[0009] Based on the acquired ship route segments, a connectivity graph of the ship's route segments between two ports is constructed, along with the characteristic information of the route segments. Combined with information on distance, arrival time, heat, and energy consumption, the optimal route is obtained.

[0010] Furthermore, cluster analysis was performed on the ship specification data based on navigability factors, including:

[0011] The ship specification data is processed using zero-mean standardization to generate data to be clustered;

[0012] The distance metric for the data to be clustered can be obtained using the following formula:

[0013]

[0014] Among them, w i x is the influence factor of the i-th attribute of the clustering object on the clustering objective. i and y i These are two sets of data to be clustered;

[0015] The clustering objects are analyzed using the Gap statistic method, and the k corresponding to the maximum value of the Gap statistic is selected as the number of clusters in this analysis.

[0016] Furthermore, the navigability factors include the year the ship was built, its length, its width, its design draft, its deadweight tonnage, and its type.

[0017] Furthermore, based on AIS data of different navigation types, vessel connectivity segments between the two ports are extracted, including:

[0018] The input AIS data is classified using MMSI numbers to obtain the continuous trajectory T = {T1, T2, ..., T...} of each vessel in the current AIS data. numMMSI};

[0019] For each consecutive trajectory, traverse through the given two port regions S. d (port1) and S d (port2) Determine if the current trajectory passes through both regions simultaneously, and finally obtain all n trajectories passing through these two ports: trajectory TS = {TS1, TS2, ..., TS}. n}

[0020] Iterate through each trajectory of TS. For the i-th trajectory of TS... i =p1, p2, ..., p n , where p i It is the i-th point of the trajectory, obtained by traversing the trajectory TS. i , obtain TS i The final m trajectories between the two ports are placed into the result set TR;

[0021] By specifying the maximum threshold md for the trajectory length between the two ports, the system filters out trajectories in TR that have excessively long paths between the two ports and returns the execution results.

[0022] Furthermore, the extraction of vessel connectivity routes between two ports based on AIS data of different navigation types also includes vessel route segment extraction based on TRACLUS, specifically including:

[0023] For each trajectory between the two ports, key points are extracted from each trajectory using MDL-based trajectory segmentation to obtain trajectory segment data based on key points, and a set of segmented line segments is obtained based on the trajectory segment data.

[0024] The optimal ε-neighborhood and MinLns are selected using entropy theory. The segment clustering algorithm is then used to cluster the processed segmented line segments according to the specified ε-neighborhood, MinLns, and MinLen values ​​to obtain a set of clustering results. The clustering results are then reassigned to the original route based on key points, and the connectivity between clusters is judged for each route.

[0025] Features are extracted from each line segment cluster to extract the entire connecting route segment between the two ports.

[0026] Furthermore, based on the acquired ship route segments, a connectivity graph of the ship's route segments between two ports is constructed, along with the characteristic information of these route segments. Combined with information on distance, arrival time, heat intensity, and energy consumption, the optimal route is obtained, including:

[0027] The objective function based on route segment planning is established as follows:

[0028] W(s i ) = w l L(s i )+w t T(s i )+w h H(s i )+w e E(s i )

[0029] Among them, w l w t w h w e These represent the proportions of distance, flight time, heat, and energy consumption in calculating the weight of the current flight segment, respectively. i T(s) is the range cost function. i H(s) is the flight time cost function. i ) is the heat cost function, E(s) i ) is the energy consumption cost function;

[0030] The optimal route is selected by calculating the weight of the connected route segments using the weight formula and filtering out the routes with the smallest weights.

[0031] Compared with the prior art, the present invention has the following advantages:

[0032] 1. This patent, based on existing AIS data, performs cluster analysis on ship navigability factors, extracts connecting route segments between two ports with different navigability types, and establishes a weight function based on route segment planning to plan ship routes in multiple dimensions such as distance, time, heat, and energy consumption. Ultimately, it designs routes that conform to the ship's own navigability factors and user preferences, which is of great significance for improving ship navigation efficiency, ensuring navigation safety, managing intelligent maritime transportation, and developing the shipping economy.

[0033] 2. This invention first proposes a route segment extraction algorithm based on ship navigability characteristics. It uses K-means++ to perform cluster analysis on ship navigability factors and improves the TRACLUS trajectory segment clustering algorithm to extract more flexible and connected route segments. Secondly, by establishing a route planning model that considers route segments, it enables route planning across multiple dimensions, including distance, time, heat, and energy consumption, allowing for more flexible route planning and better application in actual production. Experimental results show that the proposed route planning algorithm better reflects the actual navigation characteristics of ships compared to the AIS-based turning point planning algorithm. Attached Figure Description

[0034] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0035] Figure 1 This is a framework diagram of a route planning method based on AIS route segment clustering in an embodiment.

[0036] Figure 2 This is a flowchart of a route planning method based on AIS route segment clustering in an embodiment.

[0037] Figure 3 The graph shows the GAP statistics results in the example.

[0038] Figure 4 This is a graph showing the entropy values ​​of open sea areas in the example.

[0039] Figure 5 This is a graph showing the entropy values ​​of a narrow waterway in the example.

[0040] Figure 6 This is a comparison chart of the impact of navigability factors in the embodiments.

[0041] Figure 7 This is a partial trajectory map of the open sea area between the two ports extracted in the example.

[0042] Figure 8 The example shows a partial trajectory between two ports in the open sea area after classification.

[0043] Figure 9 This is a comparison chart of the planning results in the example.

[0044] Figure 10 This is a diagram showing the route planning results in the example. Detailed Implementation

[0045] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0046] like Figure 1-2 As shown, this invention provides a route planning method based on AIS route segment clustering, comprising the following steps:

[0047] S1. An algorithm for extracting vessel route segments between two ports, considering navigability factors, analyzes vessel specification data based on navigability factors to determine the navigability type standard for each vessel. Then, by extracting vessel connectivity route segments from AIS data of different navigability types, route segments that better match vessel navigation characteristics are obtained. Specifically, this includes:

[0048] S101. Clustering of ship navigability factors based on K-means++.

[0049] Based on previous research findings and the expertise of specialists in the field, the following six factors affecting ship navigation conditions have been identified.

[0050] (1) Year of Construction: Due to its age, the ship is severely corroded and has poor wind and wave resistance. If sailing in the North Pacific winter (40°C)... ° North of N, there are often northerly winds of force 10. If the ship is a slow and old vessel, the low-latitude sea area should be appropriately considered when planning the ocean route, so some detours are unavoidable.

[0051] (2-3) Length and width of the vessel: Some routes, such as straits, canals, waterways and channels for entering and leaving ports, have one or more restrictions on the length, width and height of the vessel.

[0052] (4) Design draft: Shallow-draft ships are more susceptible to wind force and are more prone to swaying, which is not conducive to the effective use of engine and rudder.

[0053] (5) Deadweight tonnage: Larger ships usually have better wind resistance.

[0054] (6) Ship type: Different types of ships carry different types of cargo, so they choose different routes.

[0055] The steps for clustering ship navigability factors based on K-means++ are as follows:

[0056] (1) Since the dimensions of the factors affecting navigation are different, in order to prevent this from affecting the clustering results, the data is processed by zero mean standardization. The calculation formula is shown in Formula 1.

[0057]

[0058] Where μ is the mean of all sample data and σ is the standard deviation of all sample data.

[0059] (2) Define the K-means++ distance metric: Since the influence of the clustered object attributes on the target result is different, in order to make the clustering result more in line with the clustering target, a weight factor is introduced on the original Euclidean distance metric to make the clustering result more in line with the expected target. The calculation formula is shown in Formula 2.

[0060]

[0061] Where w i It is the influence factor of the i-th attribute of the clustering object on the clustering target.

[0062] (3) The clustered objects are analyzed by using the Gap statistic method. The k corresponding to the maximum value of the Gap statistic is selected as the number of clusters in this case.

[0063] S102, Extraction of trajectory data between the two ports.

[0064] The route planning system studied in this invention plans routes based on given origin and destination ports. Therefore, when extracting connected route segments, it is necessary to extract the entire AIS data into ship trajectory data between the two ports in advance.

[0065] The specific steps of the extraction strategy are as follows:

[0066] (1) Classify the input AIS data by MMSI number to obtain the continuous trajectory T = {T1, T2, ..., T...} of each ship in the current AIS data. numMMSI}

[0067] (2) Since ships may not stay in the port, but may be anchored or passing through the ocean near the port, we consider the area with radius d centered on the port as the trajectory belonging to that port. Therefore, we iterate through each consecutive trajectory, passing through the given two port areas S. d (port1) and S d (port2) Determine if the current trajectory passes through both regions simultaneously. Finally, obtain all n trajectories passing through these two ports: trajectory TS = {TS1, TS2, ..., TS}. n}

[0068] (3) Since the time span of the ship's trajectory may be very long and it may make multiple round trips, we need to traverse each trajectory of TS. For the i-th trajectory of TS... i =p1, p2, ..., p n , where p i It is the i-th point of the trajectory, obtained by traversing the trajectory TS. i , obtain TS i The final m trajectories between the two ports are placed into the result set TR.

[0069] (4) Since the path between the two ports may be too long due to multiple ports, the maximum threshold md for the length of the path between the two ports is specified to filter out the path between the two ports in TR that is too long.

[0070] (5) Return the execution result and the process ends.

[0071] The pseudocode for the AIS data extraction algorithm between the two ports is shown in Algorithm 1.

[0072] S103. Extraction of ship route segments based on TRACLUS.

[0073] The detailed steps of the TRACLUS ship route segment extraction algorithm are as follows:

[0074] (1) Route segment division

[0075] a. The trajectory TR = {TR1, TR2, ..., TR} between the two ports is given by: numTra For each trajectory in}, use MDL-based trajectory partitioning for each trajectory TR i Key points are extracted to obtain key point-based trajectory segmentation data (TRP). i ={pc1pc2, pc2pc3,…,pc len-1 PC len}

[0076] b. Because the clustering results affect the number of connected trajectory segments and the connectivity between them. Using the TRACLUS algorithm's inter-trajectory metric, if the starting or ending points of two trajectory segments are spatially close, the horizontal distance between the longer and shorter segments will be very small. This leads to them being identified as directly density-reachable clustered into the same category, resulting in a decrease in the final number of trajectory segments and connectivity. Therefore, the horizontal metric in the inter-trajectory distance metric is modified to address this issue. The formula is as follows:

[0077] d || (L i L j ) = Max(l ||1 , l ||2 ,diff(L i L j (3)

[0078] c. Because the horizontal metric has been modified, longer and shorter line segments are now classified into separate categories, resulting in the loss of features in the overlapping areas of the shorter and longer segments. This reduces the flexibility of subsequent trajectory segment selection. Therefore, longer trajectories need to be further subdivided to avoid losing similarity information from shorter trajectory segments while increasing the flexibility of subsequent trajectory planning.

[0079] To avoid the smaller line segments being error segments or having very short lengths, leading to over-division of the line segments, only when the smaller line segment L... j The core line segment, i.e., |N ε (L j If the minimum segment length is greater than or equal to MinLns, and the minimum segment length is greater than the specified minimum partitioning threshold MinLen, then a partitioning decision is made. If the direct density of segment L is achievable by L... j That is, L∈N ε (L j ),|N ε (L j If the length of the segment after truncation is greater than MinLns, then truncate according to the shorter side; otherwise, ignore the segment. The final result is a set of segments D = {TRP1, TRP2, ..., TRP} with newly added feature points. numTra}

[0080] (2) Cluster the route segments and extract the connectivity between clusters.

[0081] a. Before performing line segment clustering, select the optimal ε-neighborhood and MinLns using entropy theory. When an ε-neighborhood is very small, almost all line segments L have a |N ε (L)| will all become 1; and for a very large ε neighborhood, its |N ε The value of (L)| will almost become num. ln , where num ln The total number of line segments D makes the entropy extremely large. Conversely, in good clustering, |N| ε The value of (L)| tends to slope, and the entropy decreases. Using entropy theory, we define the following formula H(X) and find the ε-neighborhood value that minimizes H(X).

[0082]

[0083]

[0084] Where n is the total number of line segment sets D, num ln .

[0085] Then, the optimal ε-neighborhood is calculated, and |N| is calculated for all line segments within this neighborhood. ε (L)| value and its corresponding avg(|N) ε (L)|). The optimal value of MinLns is finally determined to be avg(|N). ε (L)|)+C, where the constant C usually represents a range, and the specific value is ultimately determined by domain knowledge.

[0086] b. Using the line segment clustering algorithm, cluster the processed set of line segments D according to the specified ε-neighborhood, MinLns, and MinLen values ​​to obtain the set of clustering results O = {C1, C2, ..., C...} numClus The results will be saved for subsequent route segment connectivity determination.

[0087] c. Reassign the clustering results to the original key-point-based route TRPi, and perform judgment and processing on the connectivity between clusters for each route. For line segment L k ∈C y L m ∈C j If a route TRP exists i Make L k and L m If connected, then C y and C j There is a connection path between them, cluster C y Connecting C j The route is defined as N con (C y C j If |N con (C y C j If |≥MinCon, then it is called clustering C y Connectivity Clustering C j Since connectivity is determined after clustering, modifying MinCon does not require re-clustering.

[0088] (3) Extract the route segments and their characteristics.

[0089] a. For subsequent planning, it is necessary to extract features from each line segment cluster. Let H O (C i ) is cluster C i Initial heat value, V avg (C i ) is cluster C i The average transit time is calculated using formulas 6 and 7, respectively.

[0090] H O (C i )=|C i | (6)

[0091]

[0092] Where |C i |For C i The number of line segments, len(L) j ) and time(Lj ) is C i midline segment L j The length and transit time.

[0093] b. Finally, extract the entire connecting shipping route segment between the two ports. For each route, add the starting port (port1) and the ending port (port2). The sets of the left and right endpoints of its line segment cluster are respectively TRL(C i ), TRR(C i ) and TRL(C k ), TRR(C k If C i The line segment in the middle can only connect to C. k And C k The line segment in can only start from C i If they are connected, we consider these two line segments to belong to the same flight path segment, for TRR(C i ) and TRL(C k The feature points are calculated by combining the endpoints and PCR, and conversely, the feature points PCL(C) and PCR(C) are calculated separately for the two endpoint sets, as shown in Formulas 8 and 9. Finally, different routeIds are assigned as identifiers for the cluster C to which the current route segment belongs. Let RL(C) represent which cluster C can be connected to, and RR(C) represent which cluster C can be connected to. RL(C) and RR(C) are then statistically analyzed.

[0094] PCL(TRL) = center(TRL) (8)

[0095] PCR(TRR) = center(TRR) (9)

[0096] Here, center represents the center value of the endpoint set.

[0097] The pseudocode for the TRACLUS-based ship route segment extraction algorithm is shown in Algorithm 2.

[0098] S2. Construct a route planning algorithm based on route segments to solve the route.

[0099] This invention utilizes ship connectivity segments between two ports extracted from AIS data for route planning. By extracting the connectivity map of ship routes between the two ports and their characteristic information, and combining this with information on distance, arrival time, popularity, and energy consumption, a route that is both economical and reliable and reflects the actual navigation factors of ships is obtained. Based on the principles and solutions of planning problems, a ship route that meets different needs is planned. Specifically, it includes:

[0100] S201. Establishment of the objective function based on route segment planning.

[0101] Table 1. Meaning of Objective Function Symbols Based on Route Segment Planning

[0102]

[0103] (1) Route segment distance: Since the route segment is extracted from the real route, it can be navigated by ships of the same type. Therefore, the route segment distance is the distance of the route segment itself, and its formula is as follows.

[0104] L(s i ) = disi(PCL(s i ),PCR(s i (10)

[0105] Where dist(PCL(s) i ), PCR(s i )) represents line segment s i The distance between the starting point and the ending point.

[0106] (2) Flight time of a route segment: The average speed of a category of route segments can be obtained through the route segment clustering algorithm. The flight time of a route segment can be obtained through the route segment length and speed. The formula is as follows.

[0107]

[0108] (3) Route segment popularity: The number of routes in a category obtained by clustering route segments is the number of times that route segment has been selected, which is called the initial popularity H of that route segment. O Subsequently, if the route segment s i If a user selects a segment to be part of the final route, the popularity of that segment is updated using the following formula.

[0109] H(s i )=H O (s i )+H n (s i (12)

[0110] Where H n (s i The number of times the route segment was selected as the final route is denoted as .

[0111] (4) Energy consumption of the route segment: due to the sailing speed V of the route segment avgWithin the defined route segment, the resistance and thrust are constant. Therefore, when a ship navigates in oceans with similar weather conditions, resistance and thrust are constant, and the propulsion system provides all the ship's power. Its energy consumption is equal to the work done by the propulsion system to overcome the effects of the marine environment on the ship. Because the accuracy of different weather data varies and the lengths between route segments also differ, the route segment must be further divided according to the longitude of the weather data when calculating the energy consumption of a route segment. i ={sm1, sm2, ..., sm numS The formula is as follows.

[0112]

[0113] Where F env (sm j V avg (s i )), for line segment sm j The time and location of the ocean conditions affect the speed of the ship, V. avg (s i The environmental resistance generated by ).

[0114] Finally, the route segment s is obtained. i The weight function for ∈ST is as follows:

[0115] W(s i ) = w l L(s i )+w t T(s i )+w h H(s i )+w e E(s i (14)

[0116] Where w l w t w h w e These represent the weights of distance, flight time, heat, and energy consumption in calculating the weight of the current flight segment. The final weight W(s) is obtained. i ) represents route segment s i The magnitude of the travel cost, i.e., W(s) i The smaller the route segment s i The lower the cost of navigation, the better the route segment s i The easier it is to become part of the final route.

[0117] S202. Calculate the weight of the connecting route segments using the weight formula and filter out the route with the smallest weight.

[0118] The specific steps of the route planning algorithm based on route segments are as follows:

[0119] (1) Given the starting port port1 and the ending port port2, retrieve the route segment ST = {s1, s2, ..., s2} with port1 as the starting port and port2 as the ending port. numST}

[0120] (2) Given the departure time from the starting port port1 and the weights of each latitude, for Define OT(s) i ) represents route segment s i start date, LT(s) i ) represents route segment s i Forward route segment, WT(s) i The sum of the weights of the current route segment up to this route segment, including route segment s. i The weights.

[0121] (3) Starting from the starting point port1, traverse to find directly connected route segments and put them into the reachable set A.

[0122] (4) Traverse the route segments in set A and calculate the route segment with the smallest weight for each route segment. j Extract it and put it into set B and set RR(s) j If added to A, Then the path is found and the loop terminates. For Calculate LT(s) l ) = S j OT(s) l )=T(S j )+OT(S j ), WT(s l )=W(s l )+WT(S j ).

[0123] (5) Determine whether the set ST-B is empty. If it is not empty, continue to execute (4).

[0124] (6) The route with the smallest WT in port2 is the final destination and is the target route.

[0125] The pseudocode for the route planning algorithm based on route segments is shown in Algorithm 3.

[0126]

[0127]

[0128]

[0129]

[0130]

[0131]

[0132] The following specific application examples will further illustrate the solution and effects of the present invention.

[0133] This embodiment uses the DTW (Dynamic Time Warping) algorithm to measure trajectory similarity, analyzing the planned trajectory and the ship's actual navigation trajectory to ensure that the planned trajectory conforms to the ship's actual navigation characteristics.

[0134] (1) Selection of the number of clusters k.

[0135] The number of clusters k for ship navigability factors is determined by selecting a range of k values ​​from [1, 15] before clustering. For each k value within this range, the gap statistic is calculated. Figure 3 As shown.

[0136] Ultimately, the Gap statistic was maximized when the number of clusters k was 11. Therefore, this paper selects a k value of 11 for cluster analysis of ship specification data.

[0137] (2) Selection of ε-neighborhood values ​​for trajectory clustering.

[0138] The ε-neighborhood is an important parameter in trajectory clustering; its value being too large or too small will directly affect the clustering results. Since the area that ships can navigate in open sea and narrow waters is different, the trajectory density is also different. Therefore, the value of H(X) is calculated for open sea and narrow waters respectively using Formula 4 defined by entropy theory.

[0139] according to Figure 4 and Figure 5 As shown, H(X) is minimized in open sea and narrow waters when the ε-neighborhood values ​​are 1.8 and 0.7, respectively. Therefore, avg(|N) is calculated for the open sea with an ε-neighborhood value of 1.8. ε (L)|) is 3.7, and the narrow waterway avg(|N) has a value of 0.7 in the ε neighborhood. ε (L)|) is 9.3. In open sea areas, when it exceeds a certain value, the H(X) value gradually increases. The specific reason is that when the ε-neighborhood value increases, it will identify originally unrelated trajectory segments as the same type, leading to |N ε A larger (L)| value leads to a smaller number of clusters. Based on the above experimental results, this paper selects ε-neighborhood values ​​of 1.8 and 0.7 for open sea areas and narrow water areas, respectively, and MinLns values ​​of 3 and 9, respectively.

[0140] Based on the above experimental results, the hyperparameters of the model constructed in this paper are shown in Table 2.

[0141] Table 2 Hyperparameter Settings

[0142]

[0143] To verify the effectiveness of this invention in classifying ships by navigability factors for the final planning and the performance of route planning, ablation experiments were used to verify the effectiveness of ship navigability factors for the route planning model constructed in this paper in both open sea and narrow waters. The experimental results are as follows: Figure 6 As shown. The planning results of the proposed method in each dimension are calculated, along with the route planning results based on turning points. The experimental results are as follows. Figure 9 As shown.

[0144] like Figure 6 It can be seen that classifying ships in open waters significantly improves planning results, while the improvement is limited in narrow waters. This is because the navigable area for ships is smaller in narrow waters, resulting in significant similarities between the trajectories of classified ships. Consequently, the extracted route segments after classification are also similar, leading to similar planning results before and after classification. In open waters, the original trajectory of unclassified ships is as follows: Figure 7 The categorized ship trajectory map is as follows: Figure 8 It is evident that if the route is planned according to the original dimensions of distance, flight time, heat and energy consumption without classification, the final selected route segment will only be in the central part, resulting in a greater difference between the planned route and the actual route of the edge navigation, and a larger DTW value.

[0145] like Figure 9 As can be seen, the proposed method improves ship route planning in open waters to varying degrees across all four dimensions compared to route planning based on turning points, with average DTW values ​​reduced by 7.6%, 5.3%, 12.3%, and 15.8%, respectively. Therefore, compared to route planning based on turning points, the route planning based on route segments in this paper better reflects the actual course of the ship. Figure 10 Route planning based on the shortest path from turning points ignores the correlation between route points, resulting in a planned route that disregards the ship's navigation characteristics. Furthermore, routes planned using trajectory points lose characteristic information of route segments with similar spatial attributes, such as average speed and average travel time, making route planning only possible through the shortest path. Figure 10 It can also be seen that, compared with other dimensions, route planning based on heat and energy consumption in open sea areas is more in line with the navigation conditions of ships. The main reason is that, compared with other dimensions, heat and energy consumption are more in line with the ship's course preference and can dynamically adjust the planned route according to time and weather.

[0146] The improved algorithm in this invention primarily addresses the route planning problem based on AIS data. To better consider the ship's own navigational characteristics, this chapter first proposes a route segment extraction algorithm based on ship navigability features. This involves using K-means++ to perform cluster analysis on ship navigability factors and improving the TRACLUS trajectory segment clustering algorithm to extract more flexible and connected route segments. Secondly, a route planning model considering route segments is established, enabling route planning across multiple dimensions such as distance, time, heat, and energy consumption, resulting in more flexible route planning and better applicability to actual production. Experimental results show that the proposed route planning algorithm better reflects the actual navigation characteristics of ships compared to the AIS turning point-based planning algorithm.

[0147] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A route planning method based on AIS route segment clustering, characterized in that, include: Acquire vessel AIS data and vessel specification data; perform cluster analysis on vessel specification data based on navigability factors; extract vessel connectivity routes between two ports based on AIS data of different navigability types; Based on the acquired ship route segments, a connection graph of the route segments for ships sailing between two ports is constructed, along with the characteristic information of the route segments. Combined with distance, arrival time, heat, and energy consumption information, the optimal route is obtained. Based on AIS data of different navigation types, the vessel connectivity routes between the two ports are extracted, including: The input AIS data is classified using MMSI numbers to obtain the continuous trajectory of each vessel in the current AIS data. ; Traverse each consecutive trajectory, passing through the given two port areas. and Determine if the current trajectory passes through both regions simultaneously, and ultimately obtain all trajectories that pass through these two ports. Trajectory ; Traversal For each trajectory, The Trajectory ,in It is the first of the trajectories Each point is identified by traversing the trajectory. ,get The final two ports Place the trajectory into the result set middle; By specifying the maximum threshold for the trajectory length between the two ports Filter out For routes between the two ports that are too long, return the execution result; Extracting vessel connectivity routes between two ports based on AIS data of different navigation types, including vessel route segment extraction based on TRACLUS, specifically: For each trajectory between the two ports, key points are extracted from each trajectory using MDL-based trajectory segmentation to obtain trajectory segment data based on key points, and a set of segmented line segments is obtained based on the trajectory segment data. Selecting the optimal method using entropy theory Neighborhood and According to the specified Neighborhood, , The value is to use the line segment clustering algorithm to cluster the processed set of line segments to obtain a set of clustering results. The clustering results are then reassigned to the original route based on key points, and the connectivity between clusters is judged and processed for each route. Features are extracted from each line segment cluster, and then the entire connecting route segment between the two ports is extracted. In the MDL-based trajectory segmentation, the horizontal metric formula in the distance measurement between trajectories is modified as follows: in, Trajectory line segment and The horizontal distance between them is measured. and For trajectory line segments; In cases where longer and shorter line segments are classified separately after modifying the horizontal measurement, when the shorter line segment... The core line segment, i.e. And the minimum line segment length is greater than the specified minimum division threshold. When, if line segment Direct density can reach ,Right now , And the length of the cut line segment is greater than If the shortest segment is the smallest, then the segment is truncated; otherwise, the short segment is not truncated. Finally, a set of segmented segments with newly added feature points is obtained.

2. The route planning method based on AIS route segment clustering according to claim 1, characterized in that, Cluster analysis was performed on ship specification data based on navigability factors, including: The ship specification data is processed using zero-mean standardization to generate data to be clustered; The distance metric for the data to be clustered can be obtained using the following formula: in, It is the clustering object number The influence factors of each attribute on the clustering objective and These are two sets of data to be clustered; The clustered objects are analyzed using the Gap statistic method, and the value corresponding to the maximum value of the Gap statistic is selected. The number of clusters in this clustering is denoted as .

3. The route planning method based on AIS route segment clustering according to claim 2, characterized in that, The navigability factors include the year the ship was built, its length, width, design draft, deadweight tonnage, and type.

4. The route planning method based on AIS route segment clustering according to claim 1, characterized in that, Based on the acquired ship route segments, a connectivity graph of the ship's route segments between two ports is constructed, along with the characteristic information of these segments. Combined with information on distance, arrival time, heat intensity, and energy consumption, the optimal route is obtained, including: The objective function based on route segment planning is established as follows: in, , , , These represent the proportions of distance, flight time, heat, and energy consumption in calculating the weight of the current flight segment. Let the range cost function be... Let the flight time cost function be... Let the heat cost function be... Let the energy consumption cost function be used. The optimal route is selected by calculating the weight of the connected route segments using the weight formula and filtering out the routes with the smallest weights.

Images