Decision-making support method, decision-making support program, and decision-making support system

Abstract

To make it possible to quickly make a plan in which action standards opposite to a situation coexist.SOLUTION: In a decision-making support method, a pseudo integrated cost map for estimating a cost distribution is generated on an assumption that a portion corresponding to a movement route in route recording data is selected as the movement route because movement cost is the smallest. Then, by learning in which a loss function based on difference between an integrated cost map obtained by integrating a plurality of actual environment data and the pseudo integrated cost map is defined, an integrated parameter for integrating the plurality of actual environment data to generate the integrated cost map is generated. The plurality of real environment data are integrated based on the integrated parameter to generate the integrated cost map, which is displayed on a display unit.SELECTED DRAWING: Figure 6

Description

[Technical field]

[0001] The present invention relates to a decision support method, a decision support program, and a decision support system. [Background technology]

[0002] For example, a ship's navigation plan is created by a skilled operator over a long period of time, such as several days, taking into consideration a wide range of risk information, such as the marine environment, ocean condition information, target ships, fishing boat information, etc. Various techniques have been proposed to reduce the burden on operators when creating such navigation plans. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Patent Application No. 2010-237755 [Patent Document 2] International Publication No. 2018 / 163350 Summary of the Invention [Problem to be solved by the invention]

[0004] However, in the above-mentioned conventional technology, it was impossible to make a plan that had both opposing action standards, namely avoidance and approach, for situations where it was expected to avoid dangerous sea areas where pirates are infested, search for people in need of rescue, and avoid obstacles. Also, depending on the individual differences in the ability and skill of the operator, it could take a huge amount of time from planning to repeating risk analysis to making a final decision. Such circumstances apply not only to the navigation plan of a ship, but also to the planning of operation plans of various mobile machines such as cars and airplanes that move on a plane or in space while carrying people or objects.

[0005] The present invention takes the above circumstances into consideration and has as its object to enable rapid planning of a plan in which opposing standards of action coexist for a given situation. [Means for solving the problem]

[0006] In one aspect of the present invention, there is provided a decision support method executed by a decision support system that supports the determination of a movement path of a mobile machine in a real environment, the decision support system having a processor, a memory, and a storage unit, the storage unit stores a plurality of real environment data representing, for each type of information, a distribution of costs in the real environment when the movement path is determined by the mobile machine moving to each position based on the costs, and route record data recording the movement path on the map when the mobile machine moves within the real environment, and the processor reads the route record data. based on the above, generates a pseudo-integrated cost map in which the costs equal to or less than a predetermined value are distributed in a predetermined area of ​​the real environment to which the movement route in the route record data corresponds, and the costs having values ​​greater than the predetermined value are distributed in areas of the real environment other than the predetermined area; generates integration parameters for integrating the multiple real environment data to generate the integrated cost map by learning that defines a loss function based on the difference between an integrated cost map integrating the multiple real environment data and the pseudo integrated cost map; generates the integrated cost map by integrating the multiple real environment data based on the integration parameters; and displays the integrated cost map on a display unit. Effect of the Invention

[0007] According to the present invention, for example, a plan in which opposing standards of action coexist for a given situation can be quickly created. [Brief description of the drawings]

[0008] [Figure 1] FIG. 1 is an explanatory diagram of an overview of an embodiment. [Diagram 2] FIG. 1 is a diagram showing the configuration of an entire system according to a first embodiment. [Diagram 3] FIG. 1 is a diagram showing the configuration of a decision support system according to a first embodiment. [Figure 4] 4 is a flowchart showing an integrated parameter learning process according to the first embodiment. [Diagram 5] 4 is a flowchart showing a decision support process according to the first embodiment. [Figure 6] FIG. 11 is a diagram showing the configuration of a decision support system according to a second embodiment. [Figure 7] 10 is a flowchart showing a decision support process according to the second embodiment. [Figure 8] FIG. 11 is a diagram showing the configuration of a decision support system according to a third embodiment. [Figure 9] FIG. 13 is a diagram showing the configuration of a decision support system according to a fourth embodiment. [Figure 10] 13 is a flowchart showing a pseudo integrated cost map generating parameter learning process according to the fourth embodiment. [Figure 11] FIG. 13 is a diagram showing the configuration of a decision support system according to a fifth embodiment. [Figure 12] 13 is a flowchart showing an integrated parameter learning process according to the fifth embodiment. [Figure 13] FIG. 13 is a diagram showing the configuration of a decision support system according to a sixth embodiment. [Figure 14] 23 is a flowchart showing an integrated parameter learning process according to the sixth embodiment. [Figure 15] FIG. 13 is a diagram showing the configuration of a decision support system according to a seventh embodiment. [Figure 16] 13 is a flowchart showing an integrated parameter learning process according to the seventh embodiment. [Figure 17] FIG. 13 is a diagram showing the configuration of a decision support system according to an eighth embodiment. [Figure 18] FIG. 13 is a diagram showing the configuration of a decision support system according to a ninth embodiment. [Figure 19] FIG. 23 is a diagram showing a display example of the decision support system according to the ninth embodiment. [Figure 20] FIG. 2 is a diagram showing an example of the hardware configuration of a computer. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0009] Hereinafter, embodiments of the present disclosure will be described with reference to the drawings. Note that the following embodiments do not limit the present invention, and all of the elements and combinations thereof described in the following embodiments are not necessarily essential to the solution of the invention.

[0010] In the following embodiments, the following embodiments or modified examples may be described with a focus on the differences from the previously described embodiments or modified examples. In the following embodiments, the same elements or processes may be given the same reference numerals and duplicated descriptions may be omitted. In the following embodiments, the same elements or processes may be described using the same reference numerals with different branch numbers added when they are to be distinguished, and may be described using the same reference numerals without the branch numbers when they are to be collectively referred to without distinction. In the following embodiments, the same reference numerals with different branch numbers added may be used to describe similar elements or processes.

[0011] In the following embodiments, a combination of part or all of one embodiment and part or all of another embodiment within a compatible range is also included in the embodiments disclosed herein. In addition, in the following embodiments, a combination of part of an embodiment is also included in the embodiments disclosed herein.

[0012] In the following embodiments, a program that realizes each processing function unit is executed by a processor to perform a predetermined process while appropriately using a storage resource (e.g., a memory) and / or a communication interface device (e.g., a port). The processor is, for example, a CPU (Central Processing Unit). Therefore, the process executed by each processing function unit may be a process executed by a processor or a computer having the processor.

[0013] (Overview of the embodiment) FIG. 1 is an explanatory diagram of an outline of an embodiment. In the embodiment described below, the mobile machine is a ship. When the ship navigates, actual environment data D0 for each of a plurality of types of sea condition information in the marine environment or underwater environment is taken into consideration. The actual environment data D0 is, for example, map data for each type of sea condition information such as the marine environment that determines the navigation of the ship, the capabilities of the ship and the other party, and the positions of the departure and arrival points. The marine environment includes the seabed topography (water depth), ocean currents, geomagnetism, acoustic characteristics, etc. Indicators that represent the capabilities of the ship and the other party include the detection probability that the ship detects the other party and the detection probability that the ship is detected by the other party. The actual environment data D0, for example, corresponds the value of each sea condition information that determines the navigation of the ship to each position (coordinate) of a two-dimensional or three-dimensional coordinate space (actual environment) as a cost, and indicates the distribution of each cost in the coordinate space.

[0014] The integrated cost map M1 is data obtained by integrating the real-world data D0 using a parameter set of the integration parameter P0. The real-world data D0 is, for example, real-world data (ocean bottom topography) D 0,1 , Actual environment data (ocean currents) D 0,2 , Actual environment data (geomagnetic field) D 0,3 ,Acoustic characteristics of real environment data D 0,4 , .... The integrated parameter P0 is determined for each type and number of sea state information. The integrated cost map M1 is, for example, 0,i It is expressed as a linear combination of (i=1, 2, ...). However, the combination method is not limited to a linear combination.

[0015] The integrated parameter P0 is generated so as to reduce the function value of the loss function L. The integrated parameter P0 is updated, for example, by machine learning, so as to reduce the function value of the loss function L. The loss function L is, for example, a square of the result of subtracting the integrated cost map M1 from the pseudo integrated cost map M0 multiplied by a loss weight value mask M k The four arithmetic operations for each map are performed for each position (coordinate) in the coordinate space of each map.

[0016] The route record data D1 is data that records a movement route on a map that represents the actual environment when the ship moves within the same coordinate space (actual environment) as the actual environment data D0.

[0017] The pseudo integrated cost map M0 is generated so that costs below a predetermined value are distributed in a predetermined area of ​​the actual environment corresponding to the ship's movement route in the route record data D1, and costs greater than the predetermined value are distributed in areas other than this predetermined area. More specifically, for example, the distance d (or a value based on d that increases as d increases) of each position from the movement route in the map indicated by the route record data D1 is calculated. The distance of each position from the movement route is calculated using a general method for calculating the distance between a curve and a point. The pseudo integrated cost map M0 is generated using this distance d (or a value based on d that increases as d increases) as the cost at each position. In other words, the pseudo integrated cost map M0 is generated by estimating the cost distribution by interpreting that the part corresponding to the movement route in the route record data is selected as the movement route because it has the smallest movement cost.

[0018] Loss weight mask M k gives the loss weights represented by the loss function L at each location of the map, and highlights interesting regions of the map where the function values ​​of the loss function L should be considered. The loss weight mask M k is generated in a similar manner to the pseudo-integrated cost map M0. That is, the loss weight mask M k is generated so that weight values ​​equal to or greater than a predetermined value are distributed in a first region of the actual environment corresponding to the ship's movement route in the route record data D1, and weight values ​​smaller than the predetermined value are distributed in a second region other than the first region. Furthermore, the loss weight mask M k The boundary between the first and second regions may be smoothed by a Gaussian filter or the like. In addition, the loss weight mask M k may be omitted.

[0019] In the embodiment, each real environment data D 0,iThe integrated cost map M1, which integrates (i=1, 2, ...), is presented to an operator or the like who plans the ship's route. In this way, a decision support system is provided that supports the rapid planning of route plans including routes that are travel routes according to sea conditions. The decision support system also provides a decision support method.

[0020] [Embodiment 1] (Configuration of Overall System S According to First Embodiment) 2 is a diagram showing the configuration of an overall system S according to the embodiment 1. The overall system S includes a decision support system 1, an operator 2, and a ship 3. Ocean condition information data collected via a sensor 32 provided on the ship 3 and a navigation record of the ship 3 are transmitted to the decision support system 1 via a communication device 31 and used as actual environment data D0 and route record data D1.

[0021] The decision support system 1 generates an integrated cost map M1 and presents it to the operator 2. The decision support system 1 or the operator 2 determines a navigation route for the ship 3 based on the integrated cost map M1. The operator 2 steers the ship 3 via the operation unit 33 according to the determined navigation route, and navigates the ship 3 in the real environment.

[0022] (Configuration of decision support system 1 according to embodiment 1) 3 is a diagram showing a configuration of the decision support system 1 according to the embodiment 1. The decision support system 1 includes a route record data input unit 101, a route record data storage unit 102, a real environment data input unit 103, a real environment data storage unit 104, a pseudo integrated cost map generation unit 105, an integrated parameter generation unit 106, and an integrated parameter storage unit 107. The decision support system 1 also includes an integrated cost map generation unit 108 and an operator display unit 109.

[0023] The route record data input unit 101 accepts input of route record data D1 and stores it in the route record data storage unit 102. The actual environment data input unit 103 receives actual environment data D0 (actual environment data D 0,i(i=0, 1, . . . )) and stores it in the real environment data storage unit 104.

[0024] The pseudo integrated cost map generating unit 105 generates a pseudo integrated cost map M0 from the route recording data D1. The pseudo integrated cost map generating unit 105 and the integrated parameter generating unit 106 generate a pseudo integrated cost map M0-integrated cost map M1 until the learning end condition is satisfied. 2 During the learning, the integrated parameter generation unit 106 updates the integrated parameter P0 that integrates the real environment data D0, and stores the updated parameter in the integrated parameter storage unit 107.

[0025] The integrated cost map generating unit 108 generates an integrated cost map M1 by integrating the real environment data D0 for each piece of sea condition information input from the real environment data input unit 103, using the integrated parameter P0 read out from the integrated parameter storage unit 107. When generating the integrated cost map M1, the integrated cost map generating unit 108 may use the real environment data D0 stored in the real environment data storage unit 104, or may use real environment data D0 that is newer than the real environment data D0 stored in the real environment data storage unit 104.

[0026] The operator display unit 109 displays the integrated cost map M1 generated by the integrated cost map generating unit 108 and presents it to the operator 2.

[0027] (Integrated Parameter Learning Process According to the First Embodiment) 4 is a flowchart showing the integrated parameter learning process according to embodiment 1. The integrated parameter learning process according to embodiment 1 is executed by the decision support system 1 in response to an instruction from the operator 2 when learning the integrated parameter P0.

[0028] First, in step S101, the integrated parameter generating unit 106 generates an integrated parameter P wThen, in step S102, the integrated parameter generating unit 106 specifies a learning end condition. The learning end condition is the number of learning iterations, the range of the function value of the loss function L, etc.

[0029] Next, in step S103, the integrated parameter generating unit 106 generates a new integrated parameter P w to generate an integrated cost map M1.

[0030] Next, in step S104, the pseudo integrated cost map generating unit 105 reads the route record data D1 from the route record data storage unit 102, and generates a pseudo integrated cost map M0. Next, in step S105, the integrated parameter generating unit 106 compares the integrated cost map M1 with the pseudo integrated cost map M0 using the loss function L. Next, in step S106, if the result of the comparison in step S105 shows that the function value of the loss function L is smaller than that in the previous comparison, the integrated parameter generating unit 106 changes the integrated parameter P0 held in the work memory to the aforementioned new integrated parameter P w to update.

[0031] Next, in step S107, the integrated parameter generating unit 106 determines whether the learning end condition is satisfied. If the learning end condition is satisfied (step S107 YES), the integrated parameter generating unit 106 proceeds to step S108, and if the learning end condition is not satisfied (step S107 NO), the integrated parameter generating unit 106 returns to step S103.

[0032] In step S108, the integration parameter generating unit 106 saves the integration parameter P0 held in the work memory in the integration parameter storage unit 107.

[0033] (Decision-making support process according to the first embodiment) 5 is a flowchart showing the decision support processing according to embodiment 1. The decision support processing according to embodiment 1 is executed by the decision support system 1 in response to an instruction from the operator 2 when making decisions such as a route plan.

[0034] First, in step S201, the integrated cost map generating unit 108 reads out the integrated parameter P0 stored in the integrated parameter storage unit 107. Next, in step S202, the integrated cost map generating unit 108 uses the read-out integrated parameter P0 to read out and integrate the real environment data D0 stored in the real environment data storage unit 104, thereby generating an integrated cost map M1. Next, in step S203, the integrated cost map generating unit 108 displays the integrated cost map M1 generated in step S203 on the operator display unit 109.

[0035] Next, in step S208, the integrated cost map generator 108 determines whether to terminate the decision support system 1. If the decision support system 1 is to be terminated (YES in step S208), the decision support process is terminated. If the decision support system 1 is to be terminated (NO in step S208), the process returns to step S202.

[0036] (Effects of the First Embodiment) In the first embodiment, an integrated cost map integrating a plurality of real environment data is generated by learning using a loss function based on the difference between an integrated cost map integrating a plurality of real environment data and a pseudo integrated cost map based on route record data. Therefore, by presenting an integrated cost map integrating a plurality of real environment data, decision-making support that allows an operator to quickly select a route that is satisfactory to the operator can be realized. In particular, accurate and high-speed decision-making can be supported even in situations where action standards are frequently switched or where immediate decisions are required, such as in an emergency.

[0037] [Embodiment 2] The second embodiment is different from the first embodiment in that a route plan and a route analysis are performed on the integrated cost map M1 generated in the first embodiment, but otherwise is the same as the first embodiment.

[0038] (Configuration of decision support system 1B according to embodiment 2) 6 is a diagram showing a configuration of a decision support system 1B according to embodiment 2. The decision support system 1B differs from the decision support system 1 according to embodiment 1 in that it includes a route planning unit 110 and a route analysis unit 111.

[0039] The route planning unit 110 executes route planning using the Dijkstra's algorithm, the A* search algorithm, or the like, using the integrated cost map M1 generated by the integrated cost map generation unit 108 as input, and generates a planned route R0, which is a plan of a movement route of the ship 3. The route planning unit 110 displays the planned route R on the operator display unit 109.

[0040] The route analysis unit 111 calculates a profit or loss based on the cost at each position (coordinate) of the integrated cost map M1 (actual environment) that occurs when the ship 3 moves along the planned route R0, based on the planned route R0 generated by the route planning unit 110 and the actual environment data D0. The route analysis unit 111 generates route analysis information R2 that indicates the profit or loss at each calculated position (coordinate). The route analysis unit 111 displays the route analysis information R2 on the operator display unit 109.

[0041] As the route analysis information R2, for example, the result of adding up the profits and subtracting the losses obtained based on the costs at each position (coordinate) in the real environment can be used. As the route analysis information R2, the total value of the profit or loss based on the costs (risks) combined for all cost types (risk types) for each position (coordinate) in the real environment may be displayed. Alternatively, as the route analysis information R2, the respective profit or loss values ​​based on the costs (risks) for each cost type (risk type) may be displayed in a switchable manner.

[0042] (Decision-making support process according to the second embodiment) 7 is a flowchart showing the decision support process according to the embodiment 2. In step S204 following step S203 of the decision support process, the route planning unit 110 executes route planning for the integrated cost map M1 generated in step S202. Next, in step S205, the route planning unit 110 displays on the operator display unit 109 the planned route R0 generated by the route planning executed in step S204.

[0043] Next, in step S206, the route analysis unit 111 performs route analysis on the planned route R0 generated in step S204. Next, in step S207, the route analysis unit 111 displays on the operator display unit 109 the route analysis information R2 generated by the route analysis performed in step S206.

[0044] (Effects of the second embodiment) In the second embodiment, a route plan is executed using an integrated cost map as an input to generate a planned route, and the planned route is displayed on a display unit together with the integrated cost map. Since the integrated cost map has a robustness that allows accurate and fast decision-making even in situations where the action criteria are frequently switched or where an immediate decision is required, such as in an emergency, the planned route, which is the result of executing the route plan, also has a robustness. In addition, it becomes possible to provide the operator with the basis for route selection and information for route correction, and to visualize them.

[0045] [Embodiment 3] In the third embodiment, compared to the second embodiment, a behavior criterion label Lb is added as additional information when the pseudo integrated cost map generating unit 105 and the integrated parameter generating unit 106 learn the integrated parameter P0. The third embodiment is different from the second embodiment in that the integrated parameter P0 is generated for each behavior criterion indicated by the behavior criterion label Lb. The other points are the same as those of the second embodiment.

[0046] (Configuration of decision support system 1C according to embodiment 3) 8 is a diagram showing the configuration of a decision support system 1C according to embodiment 3. The decision support system 1C is different from the decision support system 1B according to embodiment 2 in that it has a behavior reference label selection unit 112 and a behavior reference label storage unit 113.

[0047] The behavioral criterion label selection unit 112 selects a behavioral criterion label Lb indicating a behavioral criterion for the ship 3 to navigate the movement route in the route record data D1 in response to an input by the operator 2. In the case of a ship, the behavioral criterion indicates a mission or purpose such as training, personnel transportation, cargo transportation, personnel and cargo transportation, and rescue. The behavioral criterion label Lb serves as identification information for the behavioral criterion.

[0048] In the third embodiment, the integrated parameter generating unit 106 learns the integrated parameter P0 using the behavior reference label Lb as additional information. The learning using the behavior reference label Lb as additional information includes (1) a first method of performing the integrated parameter learning process (FIG. 4) for each behavior reference label Lb, (2) a second method of using a Conditional-NN (Neural Net), and (3) a third method of using an E2E method.

[0049] In the first method of (1), the integrated parameter generating unit 106 performs an integrated parameter learning process (FIG. 4) for each behavior reference label Lb, and stores the behavior reference label Lb and the integrated parameter P0 in association with each other in the integrated parameter storage unit 107. The integrated cost map generating unit 108 generates an integrated cost map M1 using the integrated parameter P0 associated with the behavior reference label Lb designated by the operator 2.

[0050] In the second method of (2), the integrated parameter generating unit 106 performs the integrated parameter learning process (FIG. 4) on a Conditional-NN basis using the behavior reference label Lb as additional information. Then, a model that uses the behavior reference label Lb as input and outputs an integrated parameter P0 is learned and stored in the integrated parameter storage unit 107. The integrated cost map generating unit 108 inputs the behavior reference label Lb specified by the operator 2 into the model stored in the integrated parameter storage unit 107, and generates an integrated cost map M1 using the integrated parameter P0 output.

[0051] (Effects of the Third Embodiment) In the third embodiment, a behavioral standard label indicating a behavioral standard for a mobile machine moving along a moving route is selected according to an input by an operator, integrated parameters are generated for each behavioral standard label, and an integrated cost map is generated for each behavioral standard label. Therefore, even if the behavioral standard is changed for each action, an appropriate integrated cost map, planned route, and route analysis information can be obtained by switching the behavioral standard label.

[0052] [Embodiment 4] The fourth embodiment is different from the third embodiment in that a pseudo integrated cost map generating parameter P2 for generating a pseudo integrated cost map M0 is generated by learning. The pseudo integrated cost map M0 is generated by integrating a plurality of route recording data D1 using the pseudo integrated cost map generating parameter P2, and the rest is the same as the third embodiment.

[0053] (Configuration of decision support system 1D according to embodiment 4) 9 is a diagram showing a configuration of a decision support system 1D according to embodiment 4. The decision support system 1D differs from the decision support system 1C according to embodiment 3 in that it has a pseudo integrated cost map generation parameter learning unit 114 and a pseudo integrated cost map generation parameter storage unit 115.

[0054] The pseudo-integrated cost map generation parameter learning unit 114 learns a pseudo-integrated cost map generation parameter P2 for generating a pseudo-integrated cost map M0 based on a plurality of route recording data D1. The pseudo-integrated cost map generation parameter learning unit 114 learns a pseudo-integrated cost map generation parameter P2 for generating a pseudo-integrated cost map M0 based on a plurality of route recording data D1 from the route recording data storage unit 102. 1,i (i=1,2,...) and retrieve the route record data D 1,i Regarding the pseudo-integrated cost map M 0,i (i=1,2,…) to generate a pseudo-integrated cost map M 0,i The method of generating the pseudo integrated cost map M0 is the same as that of the pseudo integrated cost map M1.

[0055] Then, the pseudo-integrated cost map generation parameter learning unit 114 uses reinforcement learning to generate the pseudo-integrated cost map M 0,i A pseudo integrated cost map generation parameter P1 for generating a pseudo integrated cost map M0 by integrating the above is learned.

[0056] The multiple route record data D1 may be records of multiple movement routes traveled by the same ship 3 at different dates and times in the same coordinate space (actual environment) as the actual environment data D0. Alternatively, the multiple route record data D1 may be records of multiple movement routes traveled by different ships 3 in the same coordinate space (actual environment) as the actual environment data D0.

[0057] The pseudo-integrated cost map generating parameter learning unit 114 stores the generated pseudo-integrated cost map generating parameter P1 in the pseudo-integrated cost map generating parameter storage unit 115.

[0058] The pseudo integrated cost map generating unit 105 integrates a plurality of route recording data D1 read from the route recording data storage unit 102, using the pseudo integrated cost map generating parameter P1 read from the pseudo integrated cost map generating parameter storage unit 115. In this manner, the pseudo integrated cost map generating unit 105 generates the pseudo integrated cost map M0 in step S104 (FIG. 4) of the learning process of the integration parameter P0.

[0059] (Pseudo Integrated Cost Map Generation Parameter Learning Process According to the Fourth Embodiment) 10 is a flowchart showing a pseudo-integrated cost map generation parameter learning process according to embodiment 4. The pseudo-integrated cost map generation parameter learning process according to embodiment 4 is executed by the decision support system 1 in response to an instruction from the operator 2 when learning the pseudo-integrated cost map generation parameters P1.

[0060] First, in step S310, the pseudo integrated cost map generating parameter learning unit 114 initializes the pseudo integrated cost map generating parameter P1. Next, in step S302, the pseudo integrated cost map generating parameter learning unit 114 specifies a learning end condition.

[0061] Next, in step S303, the pseudo integrated cost map generating parameter learning unit 114 learns the pseudo integrated cost map generating parameter P1 by permitted learning. Next, in step S304, the pseudo integrated cost map generating parameter learning unit 114 stores the pseudo integrated cost map generating parameter P1 learned in step S303 in the pseudo integrated cost map generating parameter storage unit 115.

[0062] (Effects of the fourth embodiment) In the fourth embodiment, a pseudo-integrated cost map generating parameter for generating a pseudo-integrated cost map based on a plurality of route record data is learned, and a pseudo-integrated cost map is generated based on the plurality of route record data using the pseudo-integrated cost map generating parameter, thereby improving the robustness of the pseudo-integrated cost map and thus the robustness of the integrated cost map.

[0063] [Embodiment 5] In the fifth embodiment, compared with the second embodiment, the decision support system 1E uses a loss weight mask M k The difference is that it has a loss weight value mask generation unit 116 for generating

[0064] (Configuration of decision support system 1E according to embodiment 5) 11 is a diagram showing the configuration of a decision support system 1E according to embodiment 5. The decision support system 1E further includes a loss weight mask generating unit .

[0065] The loss weight mask generation unit 116 assigns a weight value of the loss indicated by the loss function at each position of the map represented by the difference between the integrated cost map M1 obtained by integrating a plurality of real environment data D0 and the pseudo integrated cost map M0.

[0066] The integrated parameter generation unit 106 uses the loss function L to generate a loss weight mask M k By setting the weighting coefficients of , the influence of each region in the loss function L is emphasized or weakened according to the weight to generate the integrated parameter P0. k By setting the weighting coefficient, for example, only areas of the pseudo integrated cost map M0 with a certain level of explainability or higher can be evaluated by the loss function L, and areas with low explainability can be excluded from the evaluation of the loss function L.

[0067] (Integrated Parameter Learning Process According to the Fifth Embodiment) 12 is a flowchart showing an integrated parameter learning process according to the fifth embodiment. In step S105a following step S105 of the integrated parameter learning process, the loss weight mask generating unit 116 generates a loss weight mask M based on the route record data D1 read from the route record data storage unit 102. k The areas where the explanatory power of the pseudo integrated cost map M0 is high above a certain level or low below a certain level are estimated based on the type of sea state information and the input value of the sea state information for each area of ​​the route record data D1. k is generated so that a region with higher explanatory property is assigned a larger weighting factor as a non-masked region.

[0068] Next, in step S105b, the integrated parameter generating unit 106 applies a loss weight mask M kand extracts a region with a high degree of explainability above a certain level as a region to be evaluated by the loss function L. Next, in step S106, the integrated parameter generation unit 106 compares the integrated cost map M1 with the pseudo integrated cost map M0 for the region extracted using the loss function L.

[0069] (Effects of the fifth embodiment) In the fifth embodiment, when learning the integrated parameters of the integrated cost map, a loss weight mask that gives a weight value of the loss indicated by the loss function at each position of the map is generated, and the integrated parameters are generated by learning that defines the loss function including the loss weight mask. Therefore, the presence or absence of the explainability of the pseudo integrated cost map can be reflected in the learning of the integrated parameters, and the robustness of the integrated cost map can be improved.

[0070] [Embodiment 6] The sixth embodiment is different from the second embodiment in that a decision support system 1F has an operator data designation unit 117 and a data cleansing unit 118, but is otherwise similar to the second embodiment.

[0071] (Configuration of decision support system 1F according to embodiment 6) 13 is a diagram showing the configuration of a decision support system 1F according to embodiment 6. The decision support system 1F further includes an operator data designation unit 117 and a data cleansing unit 118, as compared with the decision support system 1B of embodiment 2.

[0072] The operator data designation unit 117 receives designation of data used when the integrated parameter generation unit 106 generates the integrated parameter P0. The data designated here are the real environment data D0 read from the real environment data storage unit 104 and the route record data D1 read from the route record data storage unit 102.

[0073] The data cleansing unit 118 performs data cleansing such as selection / pruning, repair / complementation, etc. for learning according to the specified contents on the data specified by the operator data specification unit 117. The integrated parameter generation unit 106 generates an integrated parameter P0 based on the data cleansed by the data cleansing unit 118.

[0074] (Integrated Parameter Learning Process According to the Sixth Embodiment) 14 is a flowchart showing the integrated parameter learning process according to embodiment 6. In step S102a following step S102 of the integrated parameter learning process, the data cleansing unit 118 performs data cleansing.

[0075] (Effects of the Sixth Embodiment) In the sixth embodiment, data used for generating integrated parameters is cleansed based on the data designation, thereby improving the robustness of the integrated parameters.

[0076] [Embodiment 7] The seventh embodiment is different from the second embodiment in that the decision support system 1G receives a correction instruction from the operator 2 for the planned route R generated by the route planning unit 110 and displayed on the operator display unit 109, and regenerates the planned route R based on the correction instruction. The decision support system 1G also records a corrected planned route R1, which is the corrected planned route R, and generates a pseudo integrated cost map M0 based on the route recording data D1 and the corrected planned route R1. The decision support system 1G also generates a pseudo integrated cost map M0 based on the route recording data D1, the operator route R3 determined by the operator 2 from the real environment data, and the corrected planned route R1. The rest is the same as the second embodiment.

[0077] (Configuration of decision support system 1G according to embodiment 7) 15 is a diagram showing a configuration of a decision support system 1G according to embodiment 7. The decision support system 1G further includes an operator route input unit 119, an operator route storage unit 120, a planned route correction unit 121, and a corrected planned route storage unit 122, in comparison with the decision support system 1B according to embodiment 2.

[0078] The operator route input unit 119 receives an input of an operator route R3 that is determined from the actual environment data D0 and selected as an actual route by the operator 2. The operator route storage unit 120 records the operator route R3.

[0079] The planned route correction unit 121 accepts a correction instruction from the operator 2 for the planned route R generated by the route planning unit 110 and displayed on the operator display unit 109. The planned route correction unit 121 corrects the planned route R in response to the correction instruction to generate a corrected planned route R1 and stores it in the corrected planned route storage unit 122.

[0080] The pseudo integrated cost map generating unit 105 generates a pseudo integrated cost map M0 based on the route record data D1, the operator route R3, and the revised planned route R1 described below. Specifically, the pseudo integrated cost map generating unit 105 handles and processes the operator route R3 and the revised planned route R1 in the same manner as the route record data D1. Therefore, the integrated cost map M1 generated by the integrated cost map generating unit 108 at the subsequent stage reflects the operator route R3 and the revised planned route R1, and may improve the accuracy of the route planning by the route planning unit 110.

[0081] (Integrated Parameter Learning Process According to the Seventh Embodiment) 16 is a flowchart showing an integrated parameter learning process according to the seventh embodiment. In step S104a following step S104 of the integrated parameter learning process, the pseudo integrated cost map generating unit 105 generates a pseudo integrated cost map M from the revised planned route data D2 indicating the revised planned route R1. 0,2 Generate.

[0082] Next, in step S104b, the pseudo integrated cost map generating unit 105 generates a pseudo integrated cost map M 0,3 Generate.

[0083] Next, in step S104c, the integrated parameter generating unit 106 calculates the pseudo integrated cost map M 0,1 , the pseudo-integrated cost map M 0,2 , and the pseudo-integrated cost map M 0,3 The pseudo-integrated cost map M0 is generated by simply averaging or weighting the pseudo-integrated cost map M 0,1 refers to the one generated in step S104 in this embodiment.

[0084] Note that the loss weight mask M k is also generated by taking a simple average or a weighted average of the loss weight value masks generated from the route record data D1, the corrected planned route data D2, and the operator route data D3, respectively.

[0085] (Effects of the Seventh Embodiment) In the seventh embodiment, a pseudo integrated cost map is generated based on the route record data and at least one of the route actually selected by the operator and the revised planned route revised by the operator. Therefore, by including the operator's selection or revision in the data on which the integrated parameters are calculated, the robustness of the data on which the integrated parameters are calculated can be quickly improved.

[0086] [Embodiment 8] The eighth embodiment is different from the second embodiment in that a decision support system 1H receives an instruction from an operator 2 to regenerate an integrated parameter P0, and regenerates the integrated parameter P0 based on the regeneration instruction.

[0087] (Configuration of decision support system 1H according to embodiment 8) 17 is a diagram showing the configuration of a decision support system 1H according to embodiment 8. The decision support system 1H further includes an integrated parameter regeneration instructing unit 123, as compared with the decision support system 1B according to embodiment 2.

[0088] The integrated parameter regeneration instructing unit 123 receives an instruction to regenerate the integrated parameter P0 from the operator 2. The integrated parameter generating unit 106 regenerates the integrated parameter P0 based on the instruction to regenerate the integrated parameter P0 received by the integrated parameter regeneration instructing unit 123.

[0089] (Effects of the eighth embodiment) In the eighth embodiment, the integrated parameters are regenerated based on an instruction from an operator to regenerate the integrated cost map, so that the integrated cost map can be updated at any time, such as when the data on which the integrated parameters are calculated is updated or changed.

[0090] [Embodiment 9] 18 is a diagram showing the configuration of a decision support system 1H according to embodiment 9. The decision support system 1I has all the functions of the decision support systems 1 to 1H according to embodiments 1 to 8.

[0091] 19 is a diagram showing a display example of a decision support system 1H according to embodiment 9. Decision support information 1091 is displayed on an operator display unit 109 of the decision support system 1H.

[0092] The decision support information 1091 has a map display 911 , an operator route input button 912 , a display map designation button 913 , a parameter regeneration button 914 , an action criterion selection button 915 , a route correction button 916 , and a route analysis result display 917 .

[0093] The map display 911 displays the current position of the target ship 3 and the planned route R0 generated by the route planning unit 110 on a map of the real environment that is the target of route planning and route analysis. When the operator 2 operates the operator route input button 912, the operator 2 accepts the operator route R3 that has been determined by the operator 2 from the real environment data D0 and input as the actual route on the screen of the decision support information 1091.

[0094] The display map designation buttons 913 include buttons for "route planning result," "integrated risk," "risk A," "risk B," and so on.

[0095] When "Route Planning Result" is pressed, the planned route R0 generated by the route planning unit 110 is displayed on the map display 911. When "Integrated Risk" is pressed, the integrated cost map M1 integrated based on the integrated parameter P0 generated by the integrated parameter generating unit 106 is displayed on the map display 911. When "Risk A" is pressed, the real environment data D0 corresponding to "Risk A" is displayed as a map on the map display 911. When "Risk B" is pressed, the real environment data D0 corresponding to "Risk B" is displayed as a map on the map display 911.

[0096] The parameter regeneration button 914 includes buttons for "regenerate" and "specify data to be used." When the operator 2 presses "regenerate," the integrated parameter generator 106 regenerates the integrated parameter P0.

[0097] When the operator 2 presses "specify data to be used", the integrated parameter generating unit 106 accepts the specification by the operator 2 of the data (actual environment data D0 and route record data D1) to be used when generating the integrated parameter P0. Then, the data cleansing unit 118 cleanses the specified data, and the integrated parameter generating unit 106 regenerates the integrated parameter P0.

[0098] When at least one of the behavior standard selection buttons 915 is pressed by the operator 2, the selection by the operator 2 of the behavior standard label Lb indicating the behavior standard for the ship 3 to navigate the movement route in the route record data D1 is input to the behavior standard label selection unit 112.

[0099] When the route correction button 916 is operated by the operator 2, a correction instruction from the operator 2 to the planned route R generated by the route planning unit 110 and displayed on the operator display unit 109 is accepted and input to the planned route correction unit 121.

[0100] The route analysis result display 917 displays route analysis information R2 indicating the profit or loss at each position (coordinate) on the map.

[0101] (Effects of the 9th embodiment) In the ninth embodiment, the operator 2 steers the ship 3 while referring to the decision support information 1091, and new route record data D1 recorded in association with the ship steering and new actual environment data D0 recorded based on the actual environment detected by the sensor 32 (FIG. 2) are accumulated. Then, the integrated cost map M1, the planned route R0, and the route analysis information R2 are updated based on the newly accumulated route record data D1 and the actual environment data D0, thereby forming a feedback loop.

[0102] (Hardware configuration of computer 1000) 20 is a diagram showing an example of a hardware configuration of a computer 1000. The computer 1000 realizes decision support systems 1-1I including various processing function units by executing a predetermined program. The computer 1000 also realizes operator terminals connected to the decision support systems 1-1H by executing a predetermined program.

[0103] The computer 1000 includes a processor 1001 including a CPU, a main memory device 1002, an auxiliary memory device 1003, a network interface 1004, an input device 1005, and an output device 1006, all of which are interconnected via an internal communication line 1007 such as a bus.

[0104] The processor 1001 controls the overall operation of the computer 1000. The main memory device 1002 is composed of, for example, a volatile semiconductor memory, and is used as a work memory for the processor 1001. The auxiliary memory device 1003 is composed of a large-capacity nonvolatile memory device such as a hard disk device, an SSD (Solid State Drive), or a flash memory, and is used to store various programs and data for long periods of time.

[0105] An executable program 1003a stored in the auxiliary storage device 1003 is loaded into the main storage device 1002 when the computer 1000 is started up or when necessary, and is executed by the processor 1001 to realize various processing function units possessed by the decision support systems 1 to 1I.

[0106] The executable program 1003a may be recorded on a non-transitory recording medium, read from the non-transitory recording medium by a media reading device, and loaded into the main memory device 1002. Alternatively, the executable program 1003a may be obtained from an external computer via a network, and loaded into the main memory device 1002.

[0107] The network interface 1004 is an interface device for connecting the computer 1000 to each network in the system or for communicating with other computers. The network interface 1004 is, for example, configured with a NIC (Network Interface Card) for a wired LAN (Local Area Network) or a wireless LAN.

[0108] The input device 1005 is composed of a keyboard, a pointing device such as a mouse, and the like, and is used by the user to input various instructions and information to the computer 1000. The output device 1006 is composed of, for example, a display device such as a liquid crystal display, and an audio output device such as a speaker, and is used to present necessary information to the user when necessary.

[0109] Although the embodiments of the present disclosure have been described above in detail, the present disclosure is not limited to the above-described embodiments, and various modifications can be made without departing from the spirit of the present disclosure. For example, the above-described embodiments have been described in detail to clearly explain the present invention, and the present disclosure is not necessarily limited to those having all of the configurations described. In addition, it is possible to add, delete, or replace part of the configuration of the above-described embodiments with other configurations.

[0110] Furthermore, the above-mentioned configurations, functional units, processing units, etc. may be realized in part or in whole by hardware, for example, by designing them as integrated circuits. The above-mentioned configurations, functions, etc. may be realized in software by a processor interpreting and executing a program that realizes each function. Information such as the program, table, file, etc. that realizes each function can be stored in a memory, a storage device such as an HDD or SSD, or a recording medium such as an IC card, an SD card, or a DVD.

[0111] In addition, in each of the above figures, the control lines and information lines are shown as those considered necessary for the explanation, and do not necessarily show all the control lines and information lines in the actual implementation. For example, it may be considered that almost all the components are actually connected to each other.

[0112] The above-described arrangement of the processing functions and data of the decision support systems 1 to 1I is merely an example. The arrangement of the processing functions and data can be changed to an optimal arrangement in terms of the performance of the hardware and software, processing efficiency, communication efficiency, etc. [Explanation of symbols]

[0113] S: Overall system, 1, 1B, 1C, 1D, 1E, 1F, 1G, 1H, 1I: Decision support system, 101: Route record data input unit, 102: Route record data storage unit, 103: Real environment data input unit, 104: Real environment data storage unit, 105: Pseudo integrated cost map generation unit, 106: Integrated parameter generation unit, 107: Integrated parameter storage unit, 108: Integrated cost map generation unit, 109: Operator display unit, 110: Route planning unit, 111: Route analysis unit, 112: Action criterion label selection unit, 113: Action criterion label memory unit, 114: pseudo-integrated cost map generation parameter learning unit, 115: pseudo-integrated cost map generation parameter memory unit, 116: loss weight value mask generation unit, 117: operator data designation unit, 118: data cleansing unit, 119: operator path input unit, 120: operator path memory unit, 121: planned path correction unit, 122: corrected planned path memory unit, 123: integrated parameter regeneration instruction unit, 1000: computer, 1001: processor, 1002: main memory unit, 1003a: executable program.

Claims

1. A decision support method executed by a decision support system that supports determining a moving path of a mobile machine in a real environment, comprising: The decision support system includes a processor, a memory, and a storage unit; The storage unit is a plurality of real environment data sets each representing a type of information that determines the movement of the mobile machine at each position on a map that represents the real environment, the data sets representing a distribution of the costs in the real environment for each type of information when the moving route is determined by the mobile machine moving to each position based on the costs; and route record data that records the movement route on the map when the mobile machine moves within the real environment. The processor, generating a pseudo integrated cost map based on the route record data, in which the costs equal to or less than a predetermined value are distributed in a predetermined area of ​​the real environment corresponding to the movement route in the route record data, and the costs having values ​​greater than the predetermined value are distributed in an area other than the predetermined area of ​​the real environment; generating an integration parameter for integrating the plurality of real environment data to generate the integrated cost map by learning in which a loss function based on a difference between an integrated cost map obtained by integrating the plurality of real environment data and the pseudo integrated cost map is defined; generating the integrated cost map by integrating the plurality of real environment data based on the integration parameter; The integrated cost map is displayed on a display unit. A decision support method comprising each process.

2. 2. The method of claim 1 , further comprising: The processor, Executing a route plan using the integrated cost map as an input to generate a planned route which is a plan of the travel route; The planned route is displayed on the display unit together with the integrated cost map. A decision support method comprising each process.

3. 3. The method of claim 2, further comprising: The processor, generating route analysis information representing a profit or loss based on each cost incurred when the mobile machine moves along the planned route based on the planned route and the plurality of actual environment data; The route analysis information is displayed on the display unit. A decision support method comprising each process.

4. 2. The method of claim 1 , further comprising: The processor, selecting a behavior criterion label indicating a behavior criterion for the mobile machine to move along the movement route in the route record data in response to an input by an operator; generating the integrated parameters for each of the behavioral criteria labels; Generate the integrated cost map for each of the behavioral criteria labels. A decision support method comprising:

5. 3. The method of claim 2, further comprising: the mobile machine is a vessel that navigates in a marine environment or an underwater environment; The storage unit is recording, as the route record data, a route that the vessel actually navigated in the marine environment or the underwater environment as a result of the vessel maneuvering; recording the plurality of actual environment data as a plurality of types of sea state information in the marine environment or the underwater environment that are taken into consideration when navigating the ship; The processor, A route along which the ship will navigate through the marine environment or the underwater environment is generated as the planned route. A decision support method comprising:

6. 2. The method of claim 1 , further comprising: The processor, A value based on the distance of each position from the travel path on the map is calculated as the cost at each position, and the pseudo integrated cost map in which the costs are distributed is generated. A decision support method comprising:

7. 2. The method of claim 1 , further comprising: The processor, learning a pseudo-integrated cost map generation parameter for generating the pseudo-integrated cost map based on a plurality of the route record data; The pseudo-integrated cost map is generated based on a plurality of the route recording data using the pseudo-integrated cost map generation parameters. A decision support method comprising each process.

8. 2. The method of claim 1 , further comprising: The processor, generating a loss weight mask that provides a weight value of the loss indicated by the loss function at each position of the map represented by the difference; The integrated parameters are generated by learning a loss function that is obtained by multiplying the loss weight mask by a value based on the difference. A decision support method comprising:

9. 2. The method of claim 1 , further comprising: The processor, Accepting designation of data to be used when generating the integrated parameters; Cleansing the data based on the data specifications A decision support method comprising:

10. 3. The method of claim 2, further comprising: The processor, accepting an operator path determined by an operator from the real environment data; Generate the pseudo integrated cost map based on the route record data and the operator route. A decision support method comprising:

11. 3. The method of claim 2, further comprising: The processor, receiving an instruction from an operator to modify the planned route displayed on the display unit, and generating a modified planned route by modifying the planned route; Generate the pseudo integrated cost map based on the route record data and the revised planned route. A decision support method comprising:

12. 2. The method of claim 1 , further comprising: The processor, receiving an instruction from an operator to regenerate the integrated parameters; Regenerating the integration parameters based on an instruction to regenerate the integration parameters. A decision support method comprising:

13. 2. The method of claim 1 , further comprising: The processor, The real environment data is displayed on the display unit. A decision support method comprising:

14. A decision support program for causing a computer to execute the decision support method according to any one of claims 1 to 13.

15. A decision support system for supporting determination of a movement path of a mobile machine in a real environment, comprising: a real environment data storage unit that stores a plurality of real environment data representing, for each type of information, a distribution of costs in the real environment when the moving route is determined by the moving machine moving to each position on a map representing the real environment, the costs being defined as information that determines the movement of the moving machine at each position on the map representing the real environment; a route record data storage unit that stores route record data that records the movement route on the map when the mobile machine moves within the real environment; a pseudo-integrated cost map generating unit that generates a pseudo-integrated cost map based on the route record data, in which the costs equal to or less than a predetermined value are distributed in a predetermined area of ​​the real environment corresponding to the travel route in the route record data, and the costs having values ​​greater than the predetermined value are distributed in an area other than the predetermined area of ​​the real environment; an integrated parameter generation unit that generates an integrated parameter for integrating the plurality of real environment data to generate the integrated cost map by learning in which a loss function based on a difference between an integrated cost map obtained by integrating the plurality of real environment data and the pseudo integrated cost map is defined; an integrated cost map generating unit that generates the integrated cost map by integrating the plurality of real environment data based on the integration parameter; a display unit for displaying the integrated cost map; A decision support system comprising:

Images