Collision risk prediction program, collision risk prediction method and collision risk prediction device

Abstract

To provide a collision risk prediction program, a collision risk prediction method and a collision risk prediction device that reduce a risk of a collision between ships.SOLUTION: A collision risk prediction device is configured to: use time-series data of past navigation information on a ship in a predetermined region to train a machine learning model; use a trained machine learning model to predict the position of the other ship changing with time based upon time-series data of navigation information on the other ship in the predetermined region up until now; predict the position of the own ship changing with time based upon the current navigation information on the own ship in the predetermined region; generate collision risk information based upon the position of the other ship changing with time and the prediction result of the position of the own ship; and report the generated collision risk information.SELECTED DRAWING: Figure 7

Description

[Technical Field]

[0001] The present invention relates to a collision risk prediction program, a collision risk prediction method, and a collision risk prediction device. [Background technology]

[0002] In the civil engineering and construction industries, efforts are being made to reduce labor and automate work processes in response to aging populations and labor shortages. At the same time, new technologies are being actively adopted to ensure safety on-site while also reducing labor. For example, various technologies are being introduced to prevent work boats and transportation vessels from colliding with other vessels in offshore construction work.

[0003] Traditionally, when navigating a work boat or transportation vessel, the captain visually checks the surroundings and maneuvers the vessel to avoid collisions with other vessels based on judgments based on his or her own experience. Construction managers also avoid collisions by grasping the movements of other vessels based on information plotted on radar or charts, and obtaining a bird's-eye view of the area in which the work boat is moving. Some radars also have an ARPA (Automatic Radar Plotting Aids) function, which displays the closest distance and time of approach to other vessels.

[0004] Other collision avoidance technologies include the following: For example, a technology has been proposed that predicts the future state of a ship based on a predetermined schedule, predicts the future state of other ships from their current states, calculates the future collision risk between the ship and other ships based on the prediction results, and then sequentially calculates the collision risk by weighting it by time. Another proposed technology calculates the future course width of one or both of a first ship and a second ship based on the positions and past progress information of each ship, and then calculates the collision risk based on the calculated future course width. [Prior art documents] [Patent documents]

[0005] [Patent Document 1] Japanese Patent Application Publication No. 11-272999 [Patent Document 2] Japanese Patent Application Laid-Open No. 2017-182730 Summary of the Invention [Problem to be solved by the invention]

[0006] However, it is difficult for the captain to visually check or for construction managers to check the overall course of each vessel, making it difficult to properly avoid collisions between vessels. While ARPA functions are mandatory for vessels over 500 tons (commercial vessels are around 15 tons), they are not installed on all vessels, making it difficult to properly avoid collisions between vessels. Furthermore, ARPA functions geometrically calculate the routes of the vessel and other vessels as straight lines or curves with a constant curvature at a constant speed, making it difficult to avoid collisions with vessels that are accelerating, decelerating, or changing direction.

[0007] Furthermore, technology that calculates the future collision risk from the future state of one ship predicted from a schedule and the future state of other ships predicted from their current state predicts a linear course, which can deviate from actual ship maneuvering, making it difficult to appropriately avoid collisions between ships.Furthermore, technology that calculates the future course width from position and past progress information to determine collision risk does not take into account the shape of the port, which can deviate from actual ship maneuvering, making it difficult to appropriately avoid collisions between ships.

[0008] The disclosed technology has been developed in consideration of the above, and aims to provide a collision risk prediction program, a collision risk prediction method, and a collision risk prediction device that reduce the risk of collision between ships. [Means for solving the problem]

[0009] In one aspect of the collision risk prediction program, collision risk prediction method, and collision risk prediction device disclosed in the present application, a computer is caused to learn a machine learning model using time series data of past ship tracks in a specified area, and to predict the position of other ships over time using the trained machine learning model based on time series data of the positions and speeds of other ships present in the specified area up to the present, to predict the position of the ship itself over time based on the current position and speed of the ship itself in the specified area, to generate collision risk information based on the predicted results of the positions of other ships and the ship itself over time, and to notify the generated collision risk information. [Effects of the Invention]

[0010] In one aspect, the present invention can reduce the risk of collision between vessels. [Brief explanation of the drawings]

[0011] [Figure 1] FIG. 1 is a block diagram of a collision risk prediction device. [Figure 2] FIG. 2 is a diagram showing the predicted position of other ships. [Figure 3] FIG. 3 shows the collision risk assessment when departure patterns are changed. [Figure 4] FIG. 4 is a diagram showing a collision risk display screen that displays the predicted results of the ship's own position and the positions of other ships. [Figure 5] FIG. 5 is a diagram showing a collision risk display screen that displays information on the collision risk over time. [Figure 6] FIG. 6 is a diagram showing an example of a collision risk display screen in the simulation mode. [Figure 7] FIG. 7 is a diagram showing an example of a collision risk display screen in the maneuvering mode. [Figure 8] FIG. 8 is a flowchart of the collision risk prediction process in the simulation mode according to the first embodiment. [Figure 9]FIG. 9 is a flowchart of the collision risk prediction process in the ship maneuvering mode according to the first embodiment. [Figure 10] FIG. 10 shows how the collision risk display screen changes when the waypoint is changed. [Figure 11A] FIG. 11A is a flowchart of a collision risk prediction process in a vessel maneuvering mode according to the second embodiment. [Figure 11B] FIG. 11B is a flowchart of a collision risk prediction process in the vessel maneuvering mode according to the second embodiment. [Figure 12] FIG. 12 is a hardware configuration diagram of the collision risk prediction device. DETAILED DESCRIPTION OF THE INVENTION

[0012] Hereinafter, embodiments of the collision risk prediction program, collision risk prediction method, and collision risk prediction device disclosed in the present application will be described in detail with reference to the accompanying drawings. Note that the collision risk prediction program, collision risk prediction method, and collision risk prediction device disclosed in the present application are not limited to the following embodiments. [Example]

[0013] 1 is a block diagram of a collision risk prediction device. The collision risk prediction device 1 has two operation phases: a learning phase in which a machine learning model 15 performs learning, and a prediction phase in which the ship's position after a predetermined time has passed based on information about the ship and the collision risk is predicted based on the prediction result. Furthermore, in the prediction phase, the collision risk prediction device 1 has two operation modes: a simulation mode in which the collision risk is predicted when a navigation plan is created before the ship departs, and a maneuvering mode in which the collision risk is predicted when the ship is maneuvered after the ship departs.

[0014] The collision risk prediction device 1 predicts the positions of other ships present within the area in which the ship is traveling at each time, and also predicts the position of the ship at each time after the ship departs, to calculate the risk of collision between the ship and other ships. For example, the collision risk prediction device 1 is mounted on the ship. Here, the area in which the ship is traveling is set in advance as, for example, the inside of a circle with a predetermined radius centered on the ship's departure position. Hereinafter, this area will be referred to as the "target area."

[0015] The following describes in detail the collision risk prediction device 1. As shown in Fig. 1, the collision risk prediction device 1 includes a data storage unit 11, a learning execution unit 12, an information acquisition unit 13, a prediction unit 14, a machine learning model 15, a mode switching unit 16, a collision risk information generation unit 17, a display unit 18, and an input unit 19.

[0016] The data storage unit 11 has departure pattern information 111 and learning data 112. The departure pattern information 111 includes a plurality of departure patterns created by combining different departure times and ship speeds. For example, if there are seven departure times between 9:00 and 9:30, spaced apart by five minutes, and three speeds, 22 knots, 25 knots, and 30 knots, the departure pattern information 111 includes 24 departure patterns as combinations of departure times and speeds.

[0017] The learning data 112 is time-series data of past navigation information of other ships in the target area. The navigation information includes, for example, information on the position, speed, and direction of the ship at each time in the past.

[0018] The machine learning model 15 is a prediction model that receives time series data of the position, speed, and direction of a ship as input and predicts the position of the ship over time. In this embodiment, the machine learning model 15 receives time series data of the position, speed, and direction of the ship up to the present and predicts the position of the ship over time up to 60 minutes from the present. The machine learning model 15 can use a sequence to sequence model that receives time series data as input and outputs time series data.

[0019] In the learning phase, the learning execution unit 12 acquires learning data 112 from the data storage unit 11. Then, the learning execution unit 12 inputs information on the position, speed, and direction of the ship at each time, and causes the machine learning model 15 to perform learning using the positions of other ships a predetermined time after each time as training data. In this way, the learning execution unit 12 generates a trained machine learning model 15.

[0020] Here, the target area is an example of a “predetermined area.” The learning execution unit 12 then causes the machine learning model 15 to learn using time-series data of past ship navigation information in the predetermined area.

[0021] The machine learning model 15 can learn including the shape of the sea route and port by learning using the actual ship tracks in the area where the ship itself navigates. This allows the machine learning model 15 to avoid making predictions that are not actually possible, such as the presence of a ship on land, when predicting the position of the ship over time.

[0022] In this embodiment, the learning execution unit 12 that performs learning of the machine learning model 15 is mounted on the collision risk prediction device 1, but the processing of the learning execution unit 12 may be performed by another information processing device. In this case, the collision risk prediction device 1 performs the following collision risk prediction using the trained machine learning model 15 generated by the other information processing device.

[0023] The information acquisition unit 13 acquires information transmitted from an AIS (Automatic Identification System) installed on each other ship present in the target area. The AIS is a system that automatically transmits and receives information such as the identification information, type, position, course, and speed of a ship. The information acquisition unit 13 then outputs information on the identification information, position, speed, and direction of the other ships to the prediction unit 14.

[0024] The information acquisition unit 13 also acquires information about the ship from its own AIS. The information acquisition unit 13 then outputs information about the ship's position, speed, and orientation to the prediction unit 14. However, before departure, the ship is anchored at the departure point, and there is no change in the ship's position, speed, or orientation, so the information acquisition unit 13 does not need to output this information. Here, in this embodiment, AIS is used to acquire ship information, but this is just one example, and other systems can be used as long as they can acquire information about the ship, and information can also be collected using GNSS (Global Navigation Satellite System), etc.

[0025] The mode switching unit 16 receives an instruction from the operator, indicating whether the operation mode to operate in, the simulation mode or the maneuvering mode, during the prediction phase, from the input unit 19. Here, the operator refers to a person who plans the navigation plan in the simulation mode, or a person operating the ship, including the captain, in the maneuvering mode. The mode switching unit 16 then instructs the prediction unit 14 to operate in the specified operation mode.

[0026] The prediction unit 14 receives an instruction to operate in an operation mode designated by the operator from the mode switching unit 16. Then, the prediction unit 14 operates in the designated operation mode, either the simulation mode or the ship maneuvering mode.

[0027] The operation of the prediction unit 14 in the simulation mode will be described. The prediction unit 14 acquires departure pattern information 111 from the data storage unit 11. In this case, the prediction unit 14 sets the current position of the ship as the departure point. Then, for each departure pattern, the prediction unit 14 predicts the ship's position according to the elapsed time from the departure point based on the departure point, departure time, speed, and navigation plan.

[0028] For example, if waypoints are set in advance, the prediction unit 14 predicts the ship's position at each time up to the first waypoint according to the departure pattern, and then predicts the ship's position at each time at a predetermined speed after the first waypoint. There may be multiple waypoints, and the subsequent speed may change for each waypoint. Alternatively, if a higher calculation load is acceptable, the prediction unit 14 may store a speed pattern for each waypoint to the next waypoint or arrival point, and predict the ship's position according to each pattern.

[0029] The prediction unit 14 also sequentially acquires information on the position, speed, and direction of other ships present in the target area for each time up to the present from the information acquisition unit 13. The prediction unit 14 then inputs time series data on the position, speed, and direction of other ships present in the target area, which is matched to the departure time of each departure pattern, to the trained machine learning model 15. For example, the prediction unit 14 inputs information on other ships up to the departure time of each departure pattern to the trained machine learning model 15.

[0030] Thereafter, for each departure time, the prediction unit 14 acquires information on the positions of other ships over time output from the machine learning model 15. For example, the prediction unit 14 acquires information on the positions of other ships for each time from the departure time until one hour later.

[0031] FIG. 2 is a diagram showing the prediction results of the other ship's position. The prediction unit 14 acquires time-series data of the other ship's position, speed, and orientation up to the other ship's current position 201. The prediction unit 14 then inputs the acquired time-series data into the machine learning model 15, and acquires, as its output, the other ship's position, speed, and orientation over time, indicated by point 202 extending from the other ship's current position 201 in FIG. 2. Here, curve 203 indicates the actual track of the other ship. In this way, the prediction unit 14 can acquire a prediction result of a route that approximates the actual track, rather than a straight line or a curve with a predetermined curvature.

[0032] The prediction unit 14 then calculates the closest approach distance for each departure pattern using the positions of other ships and the ship itself at each time, and predicts the collision risk. For example, the prediction unit 14 can predict the collision risk using the closest approach distance and the positional relationship between the ship itself and other ships. More specifically, the prediction unit 14 predicts a lower collision risk the longer the closest approach distance, and can predict a lower collision risk when the other ship is passing the route when the closest approach distance is the same. The prediction unit 14 can also predict a lower collision risk when the other ship is in a position moving away from the ship's route. The prediction unit 14 then identifies the departure pattern with the lowest collision risk from the collision risks predicted for each departure pattern, and sets this as the recommended departure pattern. Furthermore, the prediction unit 14 calculates the closest approach distance and closest approach time between the other ships and the ship itself for each departure pattern.

[0033] Next, the prediction unit 14 outputs the recommended departure pattern, the prediction result for each departure pattern, the closest approach distance and closest approach time to the collision risk information generation unit 17. Here, in this embodiment, the prediction unit 14 outputs information on the recommended departure pattern, but this is not limited to this, and the prediction unit 14 may identify a predetermined number of departure patterns in order of lowest collision risk and output information on them.

[0034] 3 is a diagram showing an assessment of collision risk when a departure pattern is changed. The prediction unit 14 acquires prediction results 230 of the other ship positions of the other ship 211 and the own ship position of the own ship 212 for each time of a specific departure pattern. The prediction unit 14 also acquires prediction results 240 of the other ship positions and the own ship position of the own ship 212 for each time of a departure pattern that departs 20 minutes after the departure pattern of the prediction results 230.

[0035] In the prediction result 230, when the own ship 212 is at position 250, the other ship 211 is at position 251. In addition, in the prediction result 240, when the own ship 212 is at position 250, the other ship 211 is at position 252. In this case, the prediction unit 14 determines, for example, that there is no collision risk in the case of the departure pattern indicated by the prediction result 240, and determines that the departure pattern of the prediction result 240 is a departure pattern with a lower collision risk compared to the prediction pattern of the prediction result 230.

[0036] Next, the operation of the prediction unit 14 in the maneuvering mode will be described. The prediction unit 14 acquires information on the ship's position, speed, and orientation at the current time from the information acquisition unit 13. Then, the prediction unit 14 predicts the ship's position at each time from the current time onwards based on the navigation plan, using the information on the ship's position, speed, and orientation at each time up to the current time.

[0037] The prediction unit 14 also sequentially acquires information on the position, speed, and orientation of each other ship present in the target area at the current time from the information acquisition unit 13. The prediction unit 14 then inputs the time series data on the position, speed, and orientation of each other ship present in the target area to the trained machine learning model 15, and acquires information on the position of each other ship according to the elapsed time since the current time.

[0038] The prediction unit 14 then calculates the closest approach distance using the other ship's position and the ship's own position according to the time elapsed since the current time. The prediction unit 14 then outputs the closest approach distance and closest approach time to the collision risk information generation unit 17.

[0039] Alternatively, the prediction unit 14 may predict the ship's position over time after the current time when the speed is changed. Here, the prediction unit 14 may use, for example, a value obtained by adding or subtracting a predetermined amount from the current speed as the changed speed. The prediction unit 14 may then identify the speed with the lowest collision risk from the collision risks for each predicted speed, and output information on the identified speed and the prediction result to the collision risk information generation unit 17. Furthermore, the prediction unit 14 may output all prediction results when the speed is changed to the collision risk information generation unit 17.

[0040] In this way, the prediction unit 14 predicts the position of other ships over time using the trained machine learning model 15 based on time series data up to the present of navigation information of other ships present in a specified area. The prediction unit 14 also predicts the ship's position over time based on the ship's current navigation information in a specified area. The prediction unit 14 also predicts the ship's position over time before the ship departs based on the ship's current navigation information, which includes the ship's departure point, departure time, and ship speed. More specifically, the prediction unit 14 predicts the ship's position for each departure pattern that combines different departure times and ship speeds.

[0041] In the simulation mode, the collision risk information generation unit 17 receives input from the prediction unit 14 of information on the recommended departure pattern with the lowest collision risk, prediction results of the positions of other ships and the ship's own ship according to the elapsed time since the current time, the closest approach distance and the closest approach time. In the maneuvering mode, the collision risk information generation unit 17 receives input from the prediction unit 14 of prediction results of the positions of other ships and the ship's own ship according to the elapsed time since the current time, the closest approach distance, the closest approach time and the prediction results. In both the simulation mode and the maneuvering mode, the collision risk information generation unit 17 generates a collision risk display screen showing the prediction results of the positions of other ships and the ship's own ship according to the elapsed time and the collision risk, as follows:

[0042] Fig. 4 is a diagram showing a collision risk display screen that displays the predicted results of the own ship's position and the positions of other ships. For example, as shown in Fig. 4, the collision risk information generation unit 17 enters a route 311 based on the predicted result of the own ship's position of the own ship 301 and a predicted route 321 based on the predicted result of the other ship's position of the other ship 302 onto the collision risk display screen. Here, a point 312 on the route 311 of the own ship 301 represents a waypoint.

[0043] Furthermore, the collision risk information generation unit 17 places a rectangle 320 with a certain width indicating the distance that the ship will travel in a certain period of time on the collision risk display screen, in line with the route 311. Here, the collision risk information generation unit 17 generates the rectangle 320 in 40-second increments. In Figure 4, the speed of the ship 301 increases after point 312, so the length of the rectangle 320 along the route 311 becomes longer.

[0044] Furthermore, the collision risk information generation unit 17 sets a prediction circle that includes the error for each time period to indicate the error in the prediction of the other ship's position. For example, as shown in Fig. 2, the collision risk information generation unit 17 places a prediction circle 204 centered on a point 202 that indicates the other ship's predicted position at a predetermined time interval. For example, in Fig. 4, the collision risk information generation unit 17 places a prediction circle 322 on a route 321 on the collision risk display screen. By placing the prediction circle 322 in this way, the ship operator can intuitively grasp the risk visually.

[0045] FIG. 5 is a diagram showing a collision risk display screen that displays collision risk information over time. In addition to the information shown in FIG. 4, when the closest approach distance is less than a predetermined proximity threshold, the collision risk information generation unit 17 displays a collision alert on the collision risk display screen at the ship's position that is closest to the ship's position over time. Specifically, when another ship 332 approaches the ship 331 to within less than the proximity threshold at a specific time, a collision alert 333 indicating a high collision risk is displayed on the collision risk display screen. The location indicated by the collision alert 333 corresponds to the ship's position at the specific time when the ship 331 is closest to the other ship 332. Furthermore, the collision risk information generation unit 17 can also display markers 335 indicating a warning before and after the collision alert 333 on the ship's 331 route. For example, the collision risk information generation unit 17 displays the collision alert 333 indicating a high collision risk in red and the warning marker 335 in yellow.

[0046] Furthermore, the collision risk information generation unit 17 may add highlighting 336 to other ships 332 that have a high collision risk. Here, in Fig. 5, for ease of viewing, the rectangle 320 in Fig. 4 is omitted and only the width is added to the path of the own ship 331, but in this case too, the collision risk information generation unit 17 may add the rectangle 320 in the same way as in Fig. 4.

[0047] The collision risk information generator 17 may also display information indicating the closest distance 334 between the ship 331 and the other ship 332 at a specific time on the collision risk display screen. This allows the ship helmsman to intuitively grasp the degree of danger by looking at the collision risk display screen. Displaying predicted position changes over time in this way allows the ship helmsman to intuitively grasp the collision risk, which is particularly effective in the ship maneuvering mode.

[0048] In the simulation mode, the collision risk information generation unit 17 receives input of departure pattern information and generates a collision risk display screen based on the prediction results for the specified departure pattern. Furthermore, in the simulation mode, the collision risk information generation unit 17 may write information on the departure pattern with the lowest collision risk onto the collision risk display screen. This allows the ship operator to create a departure plan using visually understood information.

[0049] In addition, in the maneuvering mode, the collision risk information generation unit 17 may generate a collision risk display screen by constantly changing information indicating the ship's position and the other ships' positions over time in accordance with changes in the actual navigation of the ship and the other ships. Furthermore, the collision risk information generation unit 17 may receive input of speed change information and generate a collision risk display screen based on the prediction results corresponding to the speed change. This allows the ship operator to intuitively determine what speed change will allow risk avoidance.

[0050] After generating the collision risk display screen, the collision risk information generation unit 17 outputs the generated collision risk display screen to the display unit 18. In this way, the collision risk information generation unit 17 generates collision risk information based on the prediction results of the other ship's position and the ship's position over time. In particular, the collision risk information generation unit 17 generates, as collision risk information, a collision risk display screen showing the prediction results of the other ship's position and the ship's position over time. Furthermore, the collision risk information generation unit 17 determines that the collision risk is high when the predicted other ship's position and the ship's position are less than a predetermined proximity threshold at a specific time. Then, the collision risk information generation unit 17 displays a collision alert, which is information indicating that the collision risk is high at a specific time, on the collision risk display screen. Furthermore, the collision risk information generation unit 17 determines that the collision risk is high when the predicted other ship's position and the ship's position are less than a predetermined proximity threshold at a specific time, and displays information indicating that the collision risk is high at a specific time on the collision risk display screen. Information indicating the closest approach distance is an example of "information indicating that the collision risk is high at a specific time." In addition, in the case of the simulation mode, the collision risk information generating unit 17 generates collision risk information for each departure pattern.

[0051] The display unit 18 has a monitor or the like. The display unit 18 receives data for a collision risk display screen from the collision risk information generation unit 17. Then, the display unit 18 displays the collision risk display screen from which the data has been acquired on the monitor or the like. That is, the display unit 18 notifies the operator of the collision risk information generated by the collision risk information generation unit 17.

[0052] FIG. 6 is a diagram showing an example of a collision risk display screen in simulation mode. In FIG. 6, route 401 is the predicted result of the ship's route. Furthermore, solid lines represent routes already passed by other ships, and dashed lines represent the predicted positions of other ships. Furthermore, prediction circle 412 is information indicating the prediction error relative to position 411 where other ships will be located at a specified time. For example, by referring to FIG. 6, it can be seen that for the departure pattern used in this prediction, there are three locations on route 401 where there is a high risk of collision. Furthermore, marker 402 indicating a high collision risk shows that there is a risk of collision between route 401 and another ship located at position 411 at a specified time.

[0053] In this case, the planner of the navigation plan changes the departure pattern and causes the collision risk display screen for the changed departure pattern to be displayed on the collision risk prediction device 1. In this way, by displaying collision risk display screens for various departure patterns on the collision risk prediction device 1 and comparing them, the planner can create an optimal plan.

[0054] Figure 7 is a diagram showing an example of a collision risk display screen in maneuvering mode. In Figure 7, route 501 is the predicted result of the specified route of the ship itself. Furthermore, solid lines represent routes that other ships have already passed, and dashed lines represent the predicted results of the other ships' positions. Furthermore, prediction circle 512 is information indicating the prediction error relative to position 511, where the other ship will be located at a specified time. For example, while navigating, the ship helmsman can refer to Figure 7 to confirm that if the ship navigates route 501 at the current speed, it will come closest to the other ship located at position 511 at the point indicated by collision alert 502, and that there is a high risk of collision.

[0055] In this case, the ship navigator can change the ship's speed and display a collision risk display screen for the changed speed on the collision risk prediction device 1. In this way, by displaying collision risk display screens for various speeds on the collision risk prediction device 1 and comparing them, the ship navigator can evaluate the collision risk of the situation in real time, that is, evaluate the collision risk according to the actual situation at that time.

[0056] 8 is a flowchart of the collision risk prediction process in the simulation mode according to Example 1. Next, the flow of the collision risk prediction process in the simulation mode according to this example will be described with reference to FIG.

[0057] The prediction unit 14 acquires the departure pattern information 111 from the data storage unit 11 (step S101).

[0058] Next, the prediction unit 14 selects one departure pattern from among the plurality of departure patterns included in the departure pattern information 111 (step S102).

[0059] Next, the prediction unit 14 predicts the ship's position according to the elapsed time after the departure time using the departure point, the speed in the selected departure pattern, the departure time, and the navigation plan (step S103).

[0060] Next, the prediction unit 14 identifies other ships that exist within the target area by acquiring information on the current positions of the other ships from the information acquisition unit 13. Then, the prediction unit 14 selects one other ship from the other ships that exist within the target area (step S104).

[0061] Next, the prediction unit 14 acquires time-series data of the navigation of the selected ship, including the ship's current position, speed, and direction (step S105).

[0062] Next, the prediction unit 14 inputs time series data of the other ship's position, speed, and direction up to the present into the machine learning model 15. Then, the prediction unit 14 acquires information on the other ship's position according to the elapsed time since the departure time as its output, and predicts the other ship's position according to the elapsed time since the departure time (step S106).

[0063] The prediction unit 14 calculates the closest approach distance and closest approach time between the own ship and the other ship from the prediction results of the own ship's position and the other ship's position over time (step S107).The prediction unit 14 then outputs the prediction results of the own ship's position and the other ship's position over time, the closest approach distance and closest approach time to the collision risk information generation unit 17.

[0064] The collision risk information generation unit 17 generates a collision risk display screen by adding a display of the predicted ship position of the selected other ship to the collision risk display screen showing the ship's own position over time (step S108).

[0065] Next, the collision risk information generation unit 17 determines whether the closest approach distance between the selected other ship and the ship itself is less than the proximity threshold (step S109). If the closest approach distance is equal to or greater than the proximity threshold (step S109: No), the collision risk prediction process proceeds to step S111.

[0066] On the other hand, if the closest approach distance is less than the proximity threshold (step S109: Yes), the collision risk information generator 17 adds a collision alert to the ship's position that is the closest distance on the collision risk table screen (step S110).

[0067] Next, the prediction unit 14 determines whether or not prediction of the positions of all other ships within the target area has been completed (step S111). If there are other ships for which prediction of the positions of other ships has not been completed (step S111: No), the collision risk determination process returns to step S104.

[0068] On the other hand, if prediction of the positions of all other ships within the target area has been completed (step S111: Yes), the prediction unit 14 determines whether prediction of the own ship's position and the positions of other ships has been completed for all departure patterns (step S112).If there are any departure patterns for which prediction has not been completed (step S112: No), the collision risk determination process returns to step S102.

[0069] On the other hand, if prediction of the ship's position and other ship's positions for all departure patterns has been completed (step S112: Yes), the prediction unit 14 calculates the collision risk for each departure pattern. Then, the prediction unit 14 determines the departure pattern with the lowest collision risk as the recommended departure pattern (step S113). Thereafter, the prediction unit 14 outputs information on the recommended departure pattern to the collision risk information generation unit 17.

[0070] The collision risk information generation unit 17 outputs a collision risk display screen for each departure pattern and information on the recommended departure pattern to the display unit 18 (step S114). The display unit 18 receives input of a departure pattern designation from the navigation planner via the input unit 19 and displays the collision risk display screen for the designated departure pattern together with information on the recommended departure pattern on the monitor.

[0071] 9 is a flowchart of the collision risk prediction process in the vessel maneuvering mode according to Example 1. Next, the flow of the collision risk prediction process in the vessel maneuvering mode according to this example will be described with reference to FIG.

[0072] The prediction unit 14 acquires the current ship position, ship speed, and ship direction from the information acquisition unit 13 (step S201).

[0073] Next, the prediction unit 14 predicts the ship's position according to the elapsed time from the current time using the ship's position, ship's speed, ship's direction and navigation plan (step S202).

[0074] Next, the prediction unit 14 identifies other ships that exist within the target area by acquiring information on the current positions of the other ships from the information acquisition unit 13. Then, the prediction unit 14 selects one other ship from the other ships that exist within the target area (step S203).

[0075] Next, the prediction unit 14 acquires time-series data of the navigation of the selected ship, including the ship's current position, speed, and direction (step S204).

[0076] Next, the prediction unit 14 inputs time-series data of the other ship's current position, speed, and direction to the machine learning model 15, and obtains information on the other ship's position over time since the current time as its output. As a result, the prediction unit 14 predicts the other ship's position over time since the current time (step S205).

[0077] The prediction unit 14 calculates the closest approach distance and closest approach time between the own ship and the other ship from the prediction results of the own ship's position and the other ship's position according to the elapsed time since the current time (step S206).The prediction unit 14 then outputs the prediction results of the own ship's position and the other ship's position according to the elapsed time since the current time, the closest approach distance and closest approach time to the collision risk information generation unit 17.

[0078] The collision risk information generation unit 17 generates a collision risk display screen by adding a display of the predicted ship position of the selected other ship to the collision risk display screen showing the ship's own position according to the elapsed time since the current time (step S207).

[0079] Next, the collision risk information generation unit 17 determines whether the closest approach distance between the selected other ship and the ship itself is less than the proximity threshold (step S208). If the closest approach distance is equal to or greater than the proximity threshold (step S208: No), the collision risk prediction process proceeds to step S210.

[0080] On the other hand, if the closest approach distance is less than the proximity threshold (step S208: Yes), the collision risk information generator 17 adds a collision alert to the ship's position that is the closest distance on the collision risk table screen (step S209).

[0081] Next, the prediction unit 14 determines whether or not prediction of the positions of all other ships within the target area has been completed (step S210). If there are other ships for which prediction of the positions of other ships has not been completed (step S210: No), the collision risk determination process returns to step S203.

[0082] On the other hand, if prediction of the positions of all other ships within the target area has been completed (step S210: Yes), the collision risk information generation unit 17 outputs the generated collision risk display screen to the display unit 18 for display (step S211).

[0083] The prediction unit 14 determines whether there is an instruction to change the speed used for prediction (step S212). For example, the vessel operator refers to the collision risk display screen displayed on the monitor, and when the collision risk is high, the vessel operator inputs a new speed specification into the collision risk prediction device 1 using the input unit 19, thereby instructing to change the speed used for prediction.

[0084] If there is an instruction to change the speed (step S212: Yes), the prediction unit 14 changes the speed of the ship used for prediction to the specified speed (step S213). Then, the collision risk prediction process returns to step S202.

[0085] On the other hand, if there is no instruction to change the speed (step S212: No), the prediction unit 14 ends the collision risk prediction process.

[0086] As described above, the collision risk prediction device according to this embodiment predicts the ship's position and the positions of other ships over time according to various departure patterns before departure, and generates a collision risk display screen that displays the prediction results and collision risk. The collision risk prediction device then notifies the ship of the collision risk according to the departure pattern by providing a collision risk display screen for each departure pattern. The collision risk prediction device also notifies the ship of the recommended departure pattern with the lowest collision risk. This allows the navigation planner to visually recognize the collision risk for each departure pattern and easily determine which departure pattern is appropriate, enabling the ship planner to develop an appropriate navigation plan with reduced collision risk.

[0087] Furthermore, when maneuvering a ship after departure, the collision risk prediction device according to this embodiment predicts the ship's position and the positions of other ships over time, based on the current conditions of the ship and the other ships, and generates a collision risk display screen that displays the prediction results and the collision risk. The collision risk prediction device then predicts the ship's position over time when the speed is changed, based on the speed change, and generates a collision risk display screen that displays the prediction results and the collision risk. This allows the ship navigator to evaluate the collision risk based on the current actual situation and quickly take appropriate measures to avoid the collision risk based on the actual situation. In this way, the collision risk prediction device can contribute to reducing the risk of collision between ships both when planning a navigation plan and when maneuvering a ship, thereby enabling safe ship operation with reduced collision risk. [Example]

[0088] Next, a second embodiment will be described. A collision risk prediction device 1 according to this embodiment is also represented by the block diagram of FIG. 1. The collision risk prediction device 1 according to this embodiment changes the waypoints to change the predicted route of the ship. Details of the collision risk prediction device 1 according to this embodiment will be described below. In the following explanation, explanations of the operation of each part that is the same as in the first embodiment may be omitted.

[0089] Here, an example will be described in which the collision risk prediction device 1 operates in maneuvering mode. The display unit 18 displays time-series data on the current positions, speeds, and directions of other ships, as well as the current position, speed, and direction of the ship itself, and the predicted positions of the ship itself and other ships using the navigation plan.

[0090] The display unit 18 also stores information on points within the ship's area that can be used as candidate waypoints. When displaying the collision risk display screen, the display unit 18 also displays the candidate waypoints on the collision risk display screen in a selectable manner.

[0091] The ship helmsman can evaluate the collision risk using the collision risk display screen displayed by the display unit 18, and if the collision risk is high, he can change the speed and specify a change to the way point. Here, adding or deleting a way point is also included in the change to the way point. The ship helmsman can use an input device (not shown) to specify a way point to add from the way point candidates displayed selectably on the collision risk display screen by the display unit 18. In addition, the ship helmsman can use an input device (not shown) to specify a way point to delete from the way points on the ship's route displayed on the collision risk display screen. For example, if the current prediction result indicates a high collision risk, the ship helmsman changes the way point to avoid other ships with a high collision risk.

[0092] When a waypoint is changed, the prediction unit 14 modifies the navigation plan so that the ship passes through the specified waypoint.The prediction unit 14 then predicts the ship's position at each time from the current time using the ship's current position, speed, heading, and the modified navigation plan.The prediction unit 14 then outputs the newly predicted ship's position and other ship's positions at each time from the current time to the collision risk information generation unit 17.In this way, the prediction unit 14 receives the specification of one or more waypoints and predicts the ship's position over time based on the ship's current position, speed, and the waypoints.

[0093] The collision risk information generation unit 17 generates a collision risk display screen using the ship's position at each time from the current time onwards and the positions of other ships at each time from the current time onwards based on the navigation plan with the changed waypoints.The collision risk information generation unit 17 then outputs the collision risk display screen with the changed waypoints to the display unit 18 for display.In this way, the collision risk information generation unit 17 generates collision risk information when passing through the specified waypoints.

[0094] 10 is a diagram showing changes in the collision risk display screen when a waypoint is changed. The prediction unit 14 predicts the position of the ship 601 at each time from now on based on the current ship position, speed, direction, and navigation plan. The prediction unit 14 also inputs time-series data of other ship positions, speeds, and directions up to now into the machine learning model 15 to obtain prediction results for the position of other ship 602 at each time from now on. The collision risk information generation unit 17 generates a collision risk display screen 610 using the prediction results by the prediction unit 14.

[0095] The helmsman refers to the collision risk display screen 610 and confirms that when his ship 601 is at position 630, the other ship 602 is at position 631, and there is a high risk of collision. The helmsman then selects to add waypoint 632 from the pre-determined waypoint candidates displayed in a selectable manner.

[0096] The prediction unit 14 modifies the navigation plan so that the ship passes through the waypoint 632, and predicts the ship position of the ship 601 at each subsequent time from the present based on the current ship position, ship speed, ship direction, and the navigation plan that passes through the waypoint 632. The collision risk information generation unit 17 generates a collision risk display screen 620 using the prediction result of the ship position based on the navigation plan that passes through the waypoint 632 by the prediction unit 14.

[0097] The prediction unit 14 can also receive an instruction to change the departure time from the current location from the input unit 19, and predict the ship's position at each subsequent time from the current ship position, ship speed, ship direction, departure time, and navigation plan. In this case, the collision risk information generation unit 17 generates the collision risk display screen 620 using the prediction result of the ship's position based on the changed departure time. Furthermore, the prediction unit 14 can also receive an instruction to change both the waypoint and the departure time from the input unit 19, and predict the ship's position over time from the current ship position, ship speed, ship direction, departure time, and navigation plan that passes through the specified waypoint.

[0098] The ship helmsman refers to the collision risk display screen 620 and confirms that when the other ship 602 is located at position 631, his ship 601 will be astern of the other ship 602, and the collision risk is low. Therefore, the ship helmsman can avoid the collision risk by maneuvering his ship so that it passes through waypoint 632.

[0099] The above explanation has been made on a configuration in which the prediction unit 14 receives instructions to change the ship's speed, waypoints, and departure time, predicts the ship's position according to the time elapsed since the present in each case, and the collision risk information generation unit 17 generates a collision risk display screen for each case. In addition to this, for example, the prediction unit 14 may store information on the amount of change in speed and the combination of waypoints in advance, and automatically change either or both of the speed, the combination of waypoints, and the departure time to predict the ship's position.

[0100] In this case, the prediction unit 14 can calculate the collision risk of the ship with other ships for each route of the ship that has changed its waypoint. For example, if the collision risk with any other ship is high even if the speed is changed, or if other ships are continuously sailing the same route without any breaks, the prediction unit 14 predicts a high collision risk for that route. If the collision risk is below a predetermined risk threshold, the prediction unit 14 may notify the collision risk information generation unit 17 of the speed that has the lowest collision risk for that route as a recommended speed, and add the recommended speed to the collision risk display screen to notify the ship helmsman.

[0101] 11A and 11B are flowcharts of the collision risk prediction process in the maneuvering mode according to Example 2. Next, the flow of the collision risk prediction process in the maneuvering mode according to this example will be described with reference to Figures 11A and 11B. Here, a case will be described in which the amount of speed change and available waypoints, etc. are determined in advance, and the collision risk prediction device 1 uses these in response to a request from the helmsman to automatically suggest a speed, waypoints, and time so as to reduce the collision risk.

[0102] The prediction unit 14 acquires the current ship position, speed, and direction from the information acquisition unit 13 (step S301).

[0103] Next, if there is a waypoint that differs from the current navigation progress, the prediction unit 14 generates a navigation plan that matches the waypoint.The prediction unit 14 then predicts the ship's position over time from the current time using the ship's current position, ship's speed, ship's direction, and navigation plan, as well as the changed departure time if there is a change in the departure time (step S302).

[0104] Next, the prediction unit 14 identifies other ships that exist within the target area by acquiring information on the current positions of the other ships from the information acquisition unit 13. Then, the prediction unit 14 selects one other ship from the other ships that exist within the target area (step S303).

[0105] Next, the prediction unit 14 acquires time-series data of the navigation of the selected other ship, including the current position of the other ship, the speed of the other ship, and the direction of the other ship (step S304).

[0106] Next, the prediction unit 14 inputs time-series data of the other ship's current position, speed, and direction to the machine learning model 15, and obtains information on the other ship's position over time since the current time as its output. As a result, the prediction unit 14 predicts the other ship's position over time since the current time (step S305).

[0107] The prediction unit 14 calculates the closest approach distance and closest approach time between the own ship and the other ship from the prediction results of the own ship's position and the other ship's position according to the elapsed time since the current time (step S306).The prediction unit 14 then outputs the prediction results of the own ship's position and the other ship's position according to the elapsed time since the current time, the closest approach distance and closest approach time to the collision risk information generation unit 17.

[0108] The collision risk information generation unit 17 generates a collision risk display screen by adding a display of the predicted ship position of the selected other ship to the collision risk display screen showing the ship's own position according to the elapsed time since the current time (step S307).

[0109] Next, the collision risk information generation unit 17 determines whether the closest approach distance between the selected other ship and the ship itself is less than the proximity threshold (step S308). If the closest approach distance is equal to or greater than the proximity threshold (step S308: No), the collision risk prediction process proceeds to step S310.

[0110] On the other hand, if the closest approach distance is less than the proximity threshold (step S308: Yes), the collision risk information generator 17 adds a collision alert to the ship's position at the closest approach distance on the collision risk table screen (step S309).

[0111] Next, the prediction unit 14 determines whether or not prediction of the positions of all other ships within the target area has been completed (step S310). If there are other ships for which prediction of the positions of other ships has not been completed (step S310: No), the collision risk determination process returns to step S303.

[0112] On the other hand, if prediction of the positions of all other ships within the target area has been completed (step S310: Yes), the prediction unit 14 notifies the collision risk information generation unit 17 of the end of collision risk determination. Upon receiving the notification, the collision risk information generation unit 17 determines whether the prediction of the ship's position made by the prediction unit 14 is a prediction based on the current navigation state (step S311). Here, a prediction that is not based on the current navigation state means that any one or a combination of the speed, waypoints, and departure time used in the prediction is different from the current one.

[0113] If the prediction is based on the current navigation state (step S311: Yes), the collision risk information generating unit 17 outputs the generated collision risk display screen to the display unit 18 to display it (step S312).

[0114] The prediction unit 14 determines whether there is a request to output the recommended speed (step S313). For example, the helmsman may refer to the collision risk display screen displayed on the monitor and, if the collision risk is high, request output of the recommended speed using the input unit 19. If there is no request to output the recommended speed (step S313: Yes), the collision risk prediction process ends.

[0115] On the other hand, if the prediction is not for the current navigation state (step S311: No) or there is a request to output a recommended speed (step S313: No), the prediction unit 14 executes the following process. Here, the prediction unit 14 stores multiple values ​​of the amount of change to be added to the current speed in advance, and the prediction unit 14 sets the value obtained by adding each amount of change to the current speed as the speed to be changed. The prediction unit 14 determines whether predictions have been made for all different speeds (step S314). If there are any speeds remaining that have not been used in the prediction (step S314: No), the prediction unit 14 changes the own ship's speed to be used in the prediction to one of the speeds that have not been used in the prediction (step S315). After that, the collision risk determination process returns to step S302.

[0116] On the other hand, if the prediction is performed using all the different speeds (step S314: Yes), the prediction unit 14 calculates the collision risk of the own ship's route in that case (step S316).

[0117] Next, the prediction unit 14 determines whether the collision risk is less than the risk threshold (step S317).

[0118] If the collision risk is equal to or greater than the risk threshold (step S317: No), the prediction unit 14 determines whether prediction has been performed for all combinations of selectable way points (step S318). If there is a combination of way points for which prediction has not been performed (step S318: No), the prediction unit 14 changes the way point to match any of the combinations of way points not used in the prediction (step S319). Thereafter, the collision risk determination process returns to step S302.

[0119] If there is no combination of waypoints for which prediction has not been performed (step S318: Yes), the departure time from the current location is changed (step S320). Changing the departure time means stopping at the current location and then setting off again. Then, the collision risk determination process returns to step S302.

[0120] On the other hand, if the collision risk is equal to or lower than the risk threshold (step S317: Yes), the prediction unit 14 determines the speed with the lowest collision risk for that route as the recommended speed. The collision risk information generation unit 17 notifies the helmsman of the recommended speed by, for example, displaying the recommended speed determined by the prediction unit 14 on the display unit 18 together with a collision risk display screen, and proposes the speed with the lowest collision risk as the recommended speed (step S321). Here, if the waypoint or departure time used in the prediction for determining the recommended speed has changed, the collision risk information generation unit 17 also obtains that information from the prediction unit 14 and provides it to the user.

[0121] In this embodiment, the case of the maneuvering mode, in which it is necessary to urgently avoid the risk of collision, has been described. However, the collision risk prediction device 1 may also receive a request to change a waypoint in the simulation mode and make a prediction for a route that passes through the specified waypoint.

[0122] As described above, the collision risk prediction device according to this embodiment receives a designation for a change in waypoint, modifies the operation plan so that the ship passes through the designated waypoint, predicts the ship's position, and notifies the ship of the collision risk if the waypoint is changed. This reduces the risk of collision between ships by taking the change in waypoint into account, rather than simply changing the ship's speed, allowing ships to navigate more flexibly and safely.

[0123] (Hardware configuration) 12 is a hardware configuration diagram of the collision risk prediction device 1. Next, an example of a hardware configuration for realizing each function of the collision risk prediction device 1 will be described with reference to FIG.

[0124] 12, the collision risk prediction device 1 includes, for example, a CPU (Central Processing Unit) 91, a memory 92, a hard disk 93, a display device 94, an input device 95, and a network interface 96. The CPU 91 is connected to the memory 92, the hard disk 93, the display device 94, the input device 95, and the network interface 96 via a bus.

[0125] The display device 94 is a monitor or the like. The display device 94 realizes the function of the display unit 18 illustrated in Fig. 1. The input device 95 is a keyboard, a mouse, or the like. The input device 95 realizes the function of the input unit 19 illustrated in Fig. 1. The display device 94 and the input device 95 may be integrated into one device, such as a monitor with a touch panel.

[0126] The network interface 96 is an interface for communication between the collision risk prediction device 1 and an external device.

[0127] The hard disk 93 is an auxiliary storage device. The hard disk 93 realizes the function of the data storage unit 11 illustrated in Fig. 1. The hard disk 93 also stores various programs including programs that realize the functions of the learning execution unit 12, the information acquisition unit 13, the prediction unit 14, the machine learning model 15, the mode switching unit 16, and the collision risk information generation unit 17 illustrated in Fig. 1.

[0128] The memory 92 is a main storage device and may be, for example, a dynamic random access memory (DRAM).

[0129] The CPU 91 reads various programs from the hard disk 93, expands them into the memory 92, and executes them. As a result, the CPU 91 realizes the functions of the learning execution unit 12, the information acquisition unit 13, the prediction unit 14, the machine learning model 15, the mode switching unit 16, and the collision risk information generation unit 17 illustrated in FIG. [Explanation of symbols]

[0130] 1. Collision risk prediction device 11 Data storage unit 12 Learning Execution Department 13 Information acquisition department 14 Prediction Department 15 Machine Learning Models 16 Mode switching section 17 Collision risk information generation unit 18 Display 19 Input section 111 Departure Pattern Information 112 Training Data

Claims

1. A machine learning model is trained using time series data of past ship navigation information in a specified area, Predicting the position of other ships over time using the trained machine learning model based on time-series data of navigation information of other ships present in the specified area up to the present; Predicting the ship's position over time based on the current navigation information of the ship in the specified area; generating collision risk information based on the predicted results of the other ship's position and the ship's position over time; Notify the generated collision risk information A collision risk prediction program that causes a computer to execute processing.

2. The collision risk prediction program described in claim 1, characterized in that the process of predicting the ship's position includes a process of predicting the ship's position over time based on the ship's current navigation information, such as the ship's departure point, departure time, and ship's speed, before the ship departs.

3. The ship position prediction process includes a process of predicting the ship position for each departure pattern that is a combination of a departure time and a ship speed that are changed, The process of generating the collision risk information includes a process of generating the collision risk information for each of the departure patterns.

3. The collision risk prediction program according to claim 2.

4. the ship position prediction process includes a process of predicting the ship position over time based on the current position, speed, and the waypoints as current navigation information of the ship, in response to the designation of one or more waypoints; The process of generating the collision risk information includes a process of generating the collision risk information when passing through the waypoint.

2. The collision risk prediction program according to claim 1.

5. The collision risk prediction program described in claim 1, characterized in that the collision risk information generation process generates a collision risk display screen as the collision risk information, which shows the predicted results of the other ship's position and the ship's own position over time.

6. The collision risk prediction program described in claim 5, characterized in that the collision risk information generation process includes a process of determining that the collision risk is high when the predicted other ship position and the ship's own position are less than a predetermined proximity threshold at a specific time, and displaying information indicating that the collision risk is high at the specific time on the collision risk display screen.

7. Collision risk prediction device A machine learning model is trained using time series data of past ship navigation information in a specified area, Predicting the position of other ships over time using the trained machine learning model based on time-series data of navigation information of other ships present in the specified area up to the present; Predicting the ship's position over time based on the current navigation information of the ship in the specified area; generating collision risk information based on the predicted results of the other ship's position and the ship's position over time; Notify the generated collision risk information A collision risk prediction method comprising:

8. a learning execution unit that trains a machine learning model using time-series data of past ship navigation information in a predetermined area; a prediction unit that predicts the position of other ships over time using the trained machine learning model based on time series data of the positions of other ships present in the specified area and the navigation information of the other ships up to the present, and predicts the position of the ship itself over time based on the current navigation information of the ship itself in the specified area; a collision risk information generation unit that generates collision risk information based on the prediction results of the other ship position and the ship's own position over time by the prediction unit; and a notification unit that notifies the collision risk information generated by the collision risk information generation unit; A collision risk prediction device comprising:

Images