Device, program and method for electronic games

Abstract

To automatically determine an appropriate combination of game elements used in an electronic game even in a situation in which history information on a play by a human is not available.SOLUTION: A device of the present invention sets a game element vector indicating characteristics of a game element for each game element, and includes: a neural network 30 for outputting selection information used for determining whether or not another game element is to be combined with a game element as input data using the game element vector set for the game element included in the input data with one or a plurality of game elements as input data; and a game element selection unit 32 for selecting a game element to be combined with the game element of the input data using the selection information. A deck is automatically constituted by combining the selected game elements.SELECTED DRAWING: Figure 3

Description

[Technical Field]

[0001] The present invention relates to an apparatus, a program and a method for electronic gaming. [Background technology]

[0002] Electronic games are known in which a game is played by combining game elements (characters, cards, items, etc.) with various characteristics. In such electronic games, there is a demand for finding combinations of game elements that are frequently used. To address this demand, a method has been used in which game designers and the like suggest combinations of game elements that provide a high effect in the game based on the characteristics of the game elements. However, this method has problems such as a large processing burden on the game designers and the like. Therefore, a method has been disclosed for finding combinations of game elements that provide a high effect in the game based on the history of past play (Patent Document 1). [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Application Publication No. 2020-54776 Summary of the Invention [Problem to be solved by the invention]

[0004] However, in the technology described in Patent Document 1, machine learning is performed using historical information consisting of decks used by players in electronic games played in the past and the results of those plays. Therefore, it is not possible to propose decks that combine game elements until historical information has been accumulated. Furthermore, when new game elements are released, it is not possible to propose decks that include the new game elements. [Means for solving the problem]

[0005] One aspect of the present invention is a device that automatically combines game elements selected from a plurality of game elements used in an electronic game to propose them as a deck, and includes: a game element vector setting means that sets, for each game element, a game element vector that indicates the characteristics of that game element; a selection information output means that takes one or more of the game elements as input data and uses the game element vector set for the game elements included in the input data by the game element vector setting means to output selection information that is used to determine whether other game elements should be combined with the game element that was provided as input data; and a game element selection means that selects the game element to be combined with the game element of the input data based on the selection information, and is characterized in that the device automatically constructs a deck by combining the game elements selected by the game element selection means.

[0006] Here, it is preferable that the selection information output means is machine-learned to output the selection information indicating the game elements that will result in a high evaluation value when the electronic game is played using the game elements included in the automatically constructed deck, based on the input data.

[0007] Preferably, the selection information is the game element vector indicating the characteristics of the game element to be combined with the game element of the input data.

[0008] It is also preferable that the game element selection means refers to a game element database in which the game element vector is stored for each of the game elements, and selects the game element associated with the game element vector that is most similar to the game element vector output as the selection information as the game element to be combined with the game element of the input data.

[0009] Preferably, the selection information is a selection rate to be combined with the game element of the input data.

[0010] It is also preferable that the selection information output means outputs the selection information using an additional game element vector in addition to the game element vector set by the game element vector setting means for the game element included in the input data.

[0011] Another aspect of the present invention is a program that automatically combines game elements selected from a plurality of game elements used in an electronic game to propose them as a deck, the program causing a computer to function as: a game element vector setting means that sets, for each of the game elements, a game element vector that indicates the characteristics of that game element; a selection information output means that takes one or more of the game elements as input data and uses the game element vector set for the game elements included in the input data by the game element vector setting means to output selection information that is used to determine whether other game elements should be combined with the game element that was the input data; and a game element selection means that selects the game element to be combined with the game element of the input data based on the selection information, and the program automatically constructs a deck by combining the game elements selected by the game element selection means.

[0012] Another aspect of the present invention is a method for automatically combining game elements selected from a plurality of game elements used in an electronic game to propose them as a deck, the method comprising: causing a computer to execute a game element vector setting step for setting, for each of the game elements, a game element vector indicating the characteristics of the game element; a selection information output step for using one or more of the game elements as input data and outputting selection information used to determine whether other game elements should be combined with the game element provided as input data, using the game element vector set for the game element included in the input data by the game element vector setting step; and a game element selection step for selecting, based on the selection information, the game elements to be combined with the game elements of the input data, and automatically constructing a deck by combining the game elements selected in the game element selection step. [Effects of the Invention]

[0013] According to the present invention, it is possible to automatically determine an appropriate combination of game elements to be used in an electronic game even in a situation where there is no historical information of play by a human. Other objects of the embodiments of the present invention will become apparent by reading the entire specification. [Brief explanation of the drawings]

[0014] [Figure 1] 1 is a diagram showing the configuration of a game element automatic combination device according to an embodiment of the present invention; [Figure 2] FIG. 2 is a diagram showing the configuration of a playing tactics learning device for an electronic game according to an embodiment of the present invention. [Figure 3] FIG. 2 is a diagram showing the configuration of a game element automatic combination learning device in an embodiment of the present invention. [Figure 4] FIG. 10 is a diagram showing the configuration of a modified example of the game element automatic combination learning device in the embodiment of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0015] <Configuration of the automatic game element combination device> As shown in Fig. 1, the automatic game element combination device 100 according to the embodiment of the present invention includes a processing unit 10, a storage unit 12, an input unit 14, and an output unit 16. In other words, the automatic game element combination device 100 has the basic configuration of a computer, and can be a general PC or the like. Furthermore, some of the functions of the automatic game element combination device 100 may be realized by an external server connected via a communication means such as the Internet.

[0016] The processing unit 10 includes a means for performing arithmetic processing, such as a CPU. The processing unit 10 executes the automatic game element combination program stored in the storage unit 12 to realize the automatic game element combination process of this embodiment. The storage unit 12 includes storage means, such as a semiconductor memory or a hard disk. The storage unit 12 is accessible from the processing unit 10 and stores the automatic game element combination program, data used in the automatic game element combination process, and the like. The input unit 14 includes a means for inputting information to the automatic game element combination device 100. The input unit 14 includes, for example, a touch panel or a keyboard for receiving input from a user. The input unit 14 also includes a network interface or the like for receiving information from outside the automatic game element combination device 100, and receives the automatic game element combination program and data used in the automatic game element combination process. The output unit 16 includes a means for outputting information processed by the automatic game element combination device 100. The output unit 16 includes, for example, a display. The output unit 16 also includes a network interface or the like for transmitting information outside the automatic game element combination device 100, and transmits information to an external server or the like.

[0017] Game elements are information used in electronic games, such as characters, items, cards, etc. In electronic games, each type of game element is given information that indicates its characteristics, such as attributes and parameters (stamina value, attack power, rarity, etc.) used in the processing of the electronic game, skill activation conditions, and skill activation effects.

[0018] When playing an electronic game, a deck that combines multiple types of game elements is used. When constructing a deck, there are as many combinations as there are game elements, but there are limitations on the number of game elements that can be combined and the types of game elements that can be combined depending on the electronic game. Note that the scope of application of this embodiment is not limited to the combination of game elements in a so-called deck, but may be any combination of game elements that is used when playing an electronic game. For example, in a team-based game, the scope of application may include team formations, combinations of equipment items for characters, etc.

[0019] In this embodiment, a process for automatically determining a combination of game elements to be used in an electronic game is performed by the game element automatic combination device 100. In this embodiment, a character is used as an example of a game element, but similar processing can be performed for other game elements (items, cards, etc.).

[0020] The specific content of the electronic game does not directly affect the present invention, so a description thereof will be omitted.

[0021] The automatic game element combination device 100 functions as a play tactics learner 200 for an electronic game and an automatic game element combination learner 300 .

[0022] <Electronic game play tactics learning device> 2 shows the configuration of an electronic game playing tactics learner 200. The electronic game playing tactics learner 200 is a learner including a neural network 20 that has been trained by machine learning to receive information about the game elements that make up a deck and information about the current state of the electronic game as input, and to output actions to be taken in the electronic game in accordance with the state of the electronic game and the information about the game elements that make up the deck.

[0023] The information about game elements includes game elements included in a deck consisting of a combination of game elements, and the characteristics of the game elements that affect the electronic game. Examples of the information about game elements include the type of game element, attributes assigned to the game element, parameter values ​​in the electronic game such as hit points (HP) assigned to the game element, and special techniques assigned to the game element. However, the information is not limited to these, and may be information about game elements that affect the electronic game. The information about game elements may include all of the information about game elements that affect the electronic game, or may include a selected portion of the information about game elements that affect the electronic game.

[0024] The information relating to the game situation is information used when playing an electronic game. The information relating to the game situation may include all of the information used when playing an electronic game, or may include a selected portion of the information used when playing an electronic game.

[0025] In the case of a competitive game in which players compete using cards, information about the game situation may include, for example, game elements contained in the opponent's deck and information about those game elements, information about the competitive environment within the electronic game such as the current stage of the electronic game, and history information about past plays of the electronic game.

[0026] The neural network 20 is composed of an input layer that receives information about the game elements that make up the deck and information about the current electronic game situation, converts it into an information format that can be processed in the intermediate layer (hidden layer), and outputs it; an intermediate layer that receives output values ​​from the input layer and performs machine learning; and an output layer that outputs the output from the intermediate layer as a final output value.

[0027] The neural network 20 receives as input information about the game elements that make up the deck and information about the current electronic game situation, and outputs the next play action to be taken in the electronic game 22 based on this information. An evaluation value determined according to the result of the electronic game 22 when the action is taken in the electronic game 22 is fed back as a learning reward, and reinforcement learning is performed on the neural network 20 so that the evaluation value increases. The evaluation value is set to a higher value for a favorable situation in the electronic game, for example, according to the win rate in the electronic game 22, the score in the electronic game 22, etc.

[0028] The neural network 20 is composed of a combination of multiple layers of neurons, and machine learning is performed by updating the weighting coefficients for the input to the neurons and the activation functions for the output from the neurons according to the evaluation value, so that the neural network outputs behavior that will result in a higher evaluation value for the input.

[0029] As shown in FIG. 2, the intermediate layer of neural network 20 in this embodiment includes element vector unit 20a, which generates and outputs features for each input game element. Element vector unit 20a is trained to output a game element vector consisting of a combination of features that indicate the characteristics of each game element that makes up the input deck. For example, a game element vector consisting of N-dimensional features is output for each game element that is input as information about the game elements that make up the deck. The game element vector for a game element may be input to a subsequent network in neural network 20.

[0030] In the electronic game playing tactics learning device 200, information about the game elements constituting the deck and information about the current electronic game situation are input to the neural network 20, and the next action to be taken when playing the electronic game 22 is output. In the electronic game 22, the current electronic game situation is input to the neural network 20, and play is executed by taking the action output from the neural network 20 using the game elements constituting the deck input to the neural network 20. As a result, the situation of the electronic game 22 changes, and an evaluation value is output in accordance with the result of the play. The evaluation value is high when the result of the play is favorable. The change in the situation in the electronic game 22 is reflected as information about the current electronic game situation input to the neural network 20. Then, the neural network 20 performs machine learning so that the evaluation value increases in accordance with the input of information about the updated current electronic game situation.

[0031] By repeating this process, the electronic game playing tactics learning device 200 learns to output actions that will result in a higher evaluation value in the electronic game 22, depending on information about the game elements that make up the deck and information about the current electronic game situation.

[0032] <Game element automatic combination learning machine> 3 shows the configuration of the game element automatic combination learning device 300. The game element automatic combination learning device 300 is a learning device that includes a neural network 30 that has been trained by machine learning to receive one or more game elements as input and output selection information indicating candidates for other game elements that should be combined with the game elements to form a deck.

[0033] The neural network 30 is composed of an input layer that receives input of one or more game elements, converts them into an information format that can be processed in an intermediate layer (hidden layer), and outputs them; an intermediate layer that receives output values ​​from the input layer and performs machine learning; and an output layer that outputs the output from the intermediate layer as a final output value.

[0034] The neural network 30 receives as input one or more game elements that make up a deck, and outputs selection information for selecting other game elements that are suitable for combining with the input game elements to form a deck in the electronic game 22. Game elements are selected according to the selection information output from the neural network 30, the selected game elements are combined to form a deck, and an evaluation value determined according to the results of playing the electronic game 22 using the deck is fed back to the neural network 30 as a learning reward. Upon receiving the evaluation value feedback, the neural network 30 performs reinforcement learning on the neural network 30 so as to increase the evaluation value. As described above, the evaluation value is set to a higher value for favorable situations in the electronic game, depending on the win rate in the electronic game 22, the score in the electronic game 22, etc.

[0035] The neural network 30 is composed of a combination of multiple layers of neurons, and machine learning is performed by updating the weighting coefficients for the inputs to the neurons and the activation functions for the outputs from the neurons according to the evaluation value, so that the neural network outputs behavior that will result in a higher evaluation value for the input.

[0036] 3, in this embodiment, the intermediate layer of neural network 30 receives game element vectors generated by neural network 20 of electronic game play tactics learner 200 for each input game element. This allows automatic game element combination learner 300 to function as game element vector setting means. In other words, the game element vectors output from element vector section 20a of neural network 20 are used as inputs to the intermediate layer of neural network 30. This makes it possible to shorten the learning time in automatic game element combination learner 300.

[0037] Neural network 30 learns game element vectors by adding new game elements to the game elements input to neural network 30 in addition to the game element vectors input from electronic game play tactics learner 200. That is, element vector unit 30a of neural network 30 is trained to output a game element vector consisting of a combination of feature quantities that indicate the characteristics of each game element input to neural network 30. For example, a game element vector consisting of M-dimensional feature quantities is output for each input game element. The additionally generated game element vector (additional element vector) is input to a subsequent network in neural network 30.

[0038] The neural network 30 outputs selection information for selecting other game elements that are suitable for combining with the input game elements to form a deck in the electronic game 22. In this way, the game element automatic combination learning device 300 functions as selection information output means. In this embodiment, the selection information is a game element vector of other game elements that are suitable for combining with each other to form a deck.

[0039] The game element automatic combination learner 300 includes a game element selection unit 32. The game element selection unit 32 receives the selection information output from the neural network 30, and selects and outputs the game element represented by the game element vector that is most similar to the selection information. As a specific example, the game element selection unit 32 can access a game element database that associates all game elements that can be used in an electronic game with the game element vectors corresponding to the game elements, and refers to the game element database to select and output the game element represented by the game element vector that is most similar to the game element vector output from the neural network 30 as the selection information.

[0040] In the automatic game element combination learning device 300, one or more game elements input to the neural network 30 are combined with game elements output from the game element selection unit 32 to update the game elements input to the neural network 30. In this way, by repeating the process of adding game elements output from the game element selection unit 32 to the game elements input to the neural network 30, a deck combining the required number of game elements can be constructed.

[0041] When a deck combining the required number of game elements is completed, the game elements included in the deck are input to an electronic game play tactics learning device 200, and the actions output from the play tactics learning device 200 are applied to the electronic game 22 to play the game. An evaluation value according to the results of playing the electronic game 22 is then fed back to the neural network 30, and the neural network 30 is trained by machine learning to improve the evaluation value.

[0042] By repeating this process, the game element automatic combination learning device 300 learns to output game elements that are preferable to be combined as a deck for one or more input game elements.

[0043] If the electronic game 22 is a competitive game, it is preferable to organize two or more decks using the automatic game element combination learning device 300 and perform reinforcement learning by playing against one another using the playing tactics learning device 200.

[0044] As described above, the game element automatic combination learning device 300 in this embodiment can automatically determine appropriate combinations of game elements to be used in an electronic game even in situations where there is no historical information on human play.

[0045] In particular, because reinforcement learning, which is suitable for exploring unknown environments where the game elements and outcomes of electronic games are unknown, is applied, dynamic learning occurs even when new game elements are added, making it possible to automatically generate optimal deck formations that include new game elements. Therefore, by keeping up with the ever-increasing number of game elements due to updates to electronic games, it becomes possible to propose optimal deck formations that take new game elements into account, as well as undiscovered but strong deck formations.

[0046] <Modification> 4 shows the configuration of a modified example of a game element automatic combination learning device 400. The game element automatic combination learning device 400 is a learning device that includes a neural network 40 that has been trained by machine learning to receive one or more game elements as input and output selection information indicating candidates for other game elements that should be combined with the game elements to form a deck.

[0047] The neural network 40 is composed of an input layer that receives input of one or more game elements, converts them into an information format that can be processed in an intermediate layer (hidden layer), and outputs them; an intermediate layer that receives output values ​​from the input layer and performs machine learning; and an output layer that outputs the output from the intermediate layer as a final output value.

[0048] The neural network 40 is configured to include multiple sub-networks 40a. The sub-networks 40a are provided in the same number as the number of game elements that can be combined into a deck, and each sub-network 40a outputs selection information for one game element that indicates whether or not it is suitable to combine the input game element into a deck.

[0049] For example, the first sub-network 40a is trained to output selection information indicating whether it is suitable to combine game element A as a deck with the input game elements. The second sub-network 40a is trained to output selection information indicating whether it is suitable to combine game element B as a deck with the input game elements. Similarly, the other sub-networks 40a are trained to output selection information indicating whether it is suitable to combine different game elements as a deck with the input game elements.

[0050] Game elements are selected in accordance with the selection information output from the neural network 40, the selected game elements are combined to form a deck, and an evaluation value determined according to the results of playing the electronic game 22 using the deck is fed back to the neural network 40 as a reward for learning. Upon receiving the evaluation value feedback, the neural network 40 performs reinforcement learning on the neural network 40 so as to increase the evaluation value. The evaluation value is set, for example, according to the win rate in the electronic game 22, the score in the electronic game 22, etc.

[0051] The sub-network 40a is made up of a combination of multiple layers of neurons, and machine learning is performed by updating the weighting coefficients for the inputs to the neurons and the activation functions for the outputs from the neurons according to the evaluation value, so that the sub-network 40a outputs behavior that results in a higher evaluation value for the input.

[0052] 4, in this embodiment, the intermediate layer of sub-network 40a receives game element vectors generated by neural network 20 of electronic game playing tactics learner 200. Since sub-network 40a is configured to output selection information for each game element, it receives the game element vector of that game element from element vector unit 20a. This makes it possible to shorten the learning time in game element automatic combination learner 400.

[0053] Sub-network 40a learns a new game element vector by adding it to the game element in addition to the game element vector input from element vector unit 20a. That is, element vector unit 40b of sub-network 40a learns to output a game element vector consisting of a combination of feature quantities that indicate the characteristics of one game element that is the target of output from sub-network 40a. For example, element vector unit 40b outputs a game element vector consisting of M-dimensional feature quantities for the game element. The additionally generated game element vector (additional element vector) is input to a subsequent network in sub-network 40a.

[0054] The sub-network 40a outputs selection information indicating whether it is suitable to combine the target game element into a deck in the electronic game 22 for the input game element. In this embodiment, the selection information is a selection rate indicating whether it is suitable to combine the target game element into a deck. In other words, the higher the selection rate output from the sub-network 40a, the more suitable it is to include the game element targeted by the sub-network 40a in a deck.

[0055] Game element automatic combination learning device 400 includes game element selection unit 42. Game element selection unit 42 receives as input the selection rates output from each of sub-networks 40a, and selects and outputs the game element with the highest selection rate.

[0056] For example, if the selection rate for game element A output from the first sub-network 40a is 0.1 and the selection rate for game element B output from the second sub-network 40a is 0.2..., and the selection rate of 0.2 for game element B output from the second sub-network 40a is higher than the selection rates output from all other sub-networks 40a, then the game element selection unit 42 selects and outputs game element B.

[0057] In the automatic game element combination learning device 400, one or more game elements input to the neural network 40 are combined with game elements output from the game element selection unit 42 to update the game elements input to the neural network 40. In this way, by repeating the process of adding the game elements output from the game element selection unit 42 to the game elements input to the neural network 40, a deck combining the required number of game elements can be constructed.

[0058] When a deck combining the required number of game elements is completed, the game elements included in the deck are input to an electronic game play tactics learning device 200, and the actions output from the play tactics learning device 200 are applied to the electronic game 22 to play the game. An evaluation value according to the results of playing the electronic game 22 is then fed back to the neural network 40, and the neural network 40 is trained by machine learning to improve the evaluation value.

[0059] By repeating this process, the game element automatic combination learner 400 learns to output game elements that are preferable to be combined as a deck for one or more input game elements.

[0060] If the electronic game 22 is a competitive game, it is preferable to organize two or more decks using the automatic game element combination learning device 400 and perform reinforcement learning by playing against one another using the playing tactics learning device 200.

[0061] This technology can be used not only in electronic games but also in general combinatorial optimization, such as team organization and schedule combinations for athletes and business people, and the combination of various resources in operations research and factory plants. In such cases, it is sufficient to learn and use the individual information for resources such as athletes, personnel, and production machines as game element vectors.

[0062] As described above, the game element automatic combination learning device 400 in this embodiment can automatically determine appropriate combinations of game elements to be used in an electronic game even in situations where there is no historical information on human play. [Explanation of symbols]

[0063] 10 processing unit, 12 memory unit, 14 input unit, 16 output unit, 20 neural network, 20a element vector unit, 22 electronic game, 30 neural network, 30a element vector unit, 32 game element selection unit, 40 neural network, 40a subnetwork, 40b element vector unit, 42 game element selection unit, 100 game element automatic combination device, 200 play tactics learner, 300, 400 game element automatic combination learner.

Claims

1. 1. A device for automatically combining game elements selected from a plurality of game elements used in an electronic game to propose a deck, comprising: a game element vector setting means for setting, for each of the game elements, a game element vector that indicates a feature of the game element; a selection information output means for outputting selection information used to determine whether one or more of the game elements should be combined with the game element included in the input data, using the game element vector set by the game element vector setting means for the game element included in the input data, with respect to other game elements; a game element selection means for selecting the game element to be combined with the game element of the input data based on the selection information; Equipped with An apparatus for automatically constructing a deck by combining the game elements selected by the game element selection means.

2. 10. The apparatus of claim 1, The device is characterized in that the selection information output means has been machine-learned to output selection information indicating game elements that will result in a high evaluation value when the electronic game is played using the game elements included in the automatically constructed deck, based on the input data.

3. 3. The device according to claim 1 or 2, The device is characterized in that the selection information is the game element vector indicating characteristics of the game element to be combined with the game element of the input data.

4. 4. The apparatus of claim 3, The game element selection means refers to a game element database in which the game element vector is stored for each of the game elements, and selects the game element associated with the game element vector that is most similar to the game element vector output as the selection information as the game element to be combined with the game element of the input data.

5. 3. The device according to claim 1 or 2, The device is characterized in that the selection information is a selection rate to be combined with the game element of the input data.

6. The device according to any one of claims 1 to 5, The device is characterized in that the selection information output means outputs the selection information using an additional game element vector in addition to the game element vector set for the game element included in the input data by the game element vector setting means.

7. A program that automatically combines game elements selected from a plurality of game elements used in an electronic game to propose a deck, Computer, a game element vector setting means for setting, for each of the game elements, a game element vector that indicates a feature of the game element; a selection information output means for outputting selection information used to determine whether one or more of the game elements should be combined with the game element included in the input data, using the game element vector set by the game element vector setting means for the game element included in the input data, with respect to other game elements; a game element selection means for selecting the game element to be combined with the game element of the input data based on the selection information; and make it work, A program that automatically constructs a deck by combining the game elements selected by the game element selection means.

8. A method for automatically combining game elements selected from a plurality of game elements used in an electronic game to propose a deck, comprising: On the computer, a game element vector setting step of setting, for each of the game elements, a game element vector that indicates a feature of the game element; a selection information output step of using one or more of the game elements as input data and using the game element vectors set for the game elements included in the input data in the game element vector setting step to output selection information used to determine whether other game elements should be combined with the game element set as the input data; a game element selection step of selecting the game element to be combined with the game element of the input data based on the selection information; Execute A method comprising automatically constructing a deck by combining the game elements selected in the game element selection step.

Images