KR20220061290A - A apparatus and a method for generating game maps - Google Patents

Abstract

Disclosed are an apparatus and a method for automatically generating game maps having features similar to those of a sample map by using the sample map and normalized data about various features of multiple game maps. The apparatus for generating game maps, which is a game map generator that generates game maps with similar features by using a generative adversarial network (GAN), comprises: a generative network receiving a sample game map as an input and trained to generate an analogous map having features similar to those of the input game map; a discriminative network trained to evaluate the analogous map using the sample game map and a game map feature data set related to features of multiple game maps; and a suitability judgment unit determining a game's suitability with the analogous map generated by the generative network.

Description

Apparatus and method for generating a computer game map

The present invention relates to an apparatus and method for generating a computer game map, and more particularly, to generate a game map similar to the sample map, based on the sample map, using a generative adverserial network (GAN). It relates to an apparatus and method for

As the computer game industry grows, various types of computer games are being developed and serviced. Most computer games are configured to receive a command from a user through a user interface such as a keyboard, mouse, and touch screen connected to a computer device and manipulate objects in the game according to the user input. In this case, a game map is used as a virtual space in which a computer game is performed.

The game map is a space in which objects necessary for the game are arranged, and the user tries to achieve a given goal by manipulating, using, avoiding, destroying, etc., the objects arranged on the game map.

Even in the same game, the user's manipulation or strategy to achieve the game goal may vary depending on the game map configuration, so the game map design is an important factor that greatly affects the user's interest in the game.

In addition, in order to maintain users' interest even after a game is released, a new game map may be additionally provided, or an existing map may need to be modified or expanded.

Recently, with the development of computer technology and data communication, more complex and various types of games have been released than before. Since the complexity and diversity of the game causes the complexity and diversity of the game map, which is the space in which the game is performed, it takes considerable time and money to design the game map.

Existing game maps can be referenced as a way to reduce the time and cost required for game map design, but there are too many problems with released games and game maps used for them. Therefore, in order to refer to the existing game map, it is first necessary to check which of the existing game maps is suitable for the characteristics of the game to be developed.

On the other hand, it is necessary to provide a variety of game maps for one game, and it is preferable that game maps with similar characteristics are provided according to the theme or difficulty of the mission. It takes considerable time and effort to produce a plurality of similar game maps, and the conventional manual map production method by a map designer is very inefficient in producing a plurality of similar maps.

SUMMARY OF THE INVENTION The present invention is to solve the problems of the prior art, and is an apparatus for automatically producing a game map having characteristics similar to a sample map by using data and a sample map obtained by normalizing data related to various characteristics of a plurality of computer game maps and methods.

A game map generating apparatus according to an embodiment of the present invention is a game map generating apparatus for generating a game map having similar characteristics using a generative adversarial network (GAN), and by receiving a sample game map as an input, the Evaluate the similar map using a generative network trained to generate a similar map having characteristics similar to those of the game map, a game map characteristic data set related to the characteristics of a plurality of game maps, and the sample game map and a discriminative network trained to do so, and a suitability determining unit for determining game suitability of a similar map generated by the generating network.

the generation network determines one or more symmetry axes from the arrangement of map tiles constituting the sample map, and randomly shuffles objects to be placed on the map tiles with respect to a first area among a plurality of areas divided by the symmetry axes; It may be configured to mirror the object shuffled with respect to the first region to another region with respect to the axis of symmetry.

The determination network compares the distribution of the game map characteristic data set and the similarity map to determine a first degree of similarity of the similarity map, compares the sample map characteristic with the similarity map characteristic, and compares the similarity map characteristic. and determine a second degree of similarity of the characteristic of the map, and evaluate the similarity map according to the first degree of similarity and the second degree of similarity. In this case, when the first and second similarities reach the first and second thresholds, respectively, the learning of the generation network may be terminated.

The game map characteristic data includes reference characteristic data regarding a reference characteristic of each map, and one or more auxiliary characteristic data regarding an auxiliary characteristic of each map,

The reference characteristic may be divided into a plurality of reference attributes, and the reference characteristic data and the auxiliary characteristic data may be mapped in a 3D space.

The mapping to the three-dimensional space includes setting reference coordinates of each of the plurality of reference attributes in a three-dimensional space using the reference characteristic data of a game map, and fitting the auxiliary characteristic data based on the reference coordinates It can be done by

The setting of the reference coordinates of each of the reference properties includes obtaining a first eigenvector in which the variance of the reference property vector with respect to a plurality of game maps is maximally preserved, and a plurality of eigenvectors orthogonal to the first eigenvector, Including the first eigenvector, three eigenvectors are selected from among the plurality of eigenvectors in the order of increasing eigenvalues to be determined as a first eigenvector group, and the first eigenvector group is divided into three axes. This can be performed by setting the reference coordinates of each reference property in a three-dimensional space of .

The fitting of the auxiliary feature data to the reference coordinates includes obtaining a second eigenvector in which the variance of the auxiliary feature vector is maximally preserved, and a plurality of eigenvectors orthogonal to the second eigenvector, and the second eigenvector including, selecting three eigenvectors from among the plurality of eigenvectors in the order of increasing eigenvalues to determine a second eigenvector group, and a three-dimensional space using the second eigenvector group as three axes This may be performed by projecting an auxiliary feature vector to , matching the directions of the first eigenvector group and the second eigenvector group, and arranging the mapping of the auxiliary feature vector with respect to the reference coordinates.

According to an embodiment of the present invention, the reference characteristic may be a type of mission performed to clear a computer game map. At this time, the types of missions are N (a natural number where N>3), and the reference characteristic data may be an N-bit multi-hot encoded bit stream in which whether each mission is relevant is coded as 0 or 1.

According to an embodiment of the present invention, the auxiliary characteristic data includes data regarding a combination of directions that the user must attack in order to clear the computer game map (hereinafter, “attack direction data”) and data regarding the initial arrangement of objects ( Hereinafter, "object initial arrangement data") may be included.

The object initial arrangement data may be an m×n matrix in which 0 or 1 is coded to indicate whether an object is arranged on each map tile constituting the map.

On the other hand, the game map generating method according to an embodiment of the present invention is a game map generating method for generating a game map with similar characteristics using a generative adversarial network (GAN), the learning and similarity of the GAN generating a distribution of characteristic data of an actual game map used for determination; receiving a sample game map for learning as an input; generating a similar map similar to distribution of characteristic data of the actual game map; and distribution of characteristic data of the actual game map of the plurality of Training the generation network and the discrimination network of the GAN to evaluate the similar map using the game map characteristic data set and the sample game map related to the characteristics of the game map; generating a similar map similar to the map, and determining whether the similar map generated by the generating network is suitable for an actual game.

The step of learning and the step of generating the similar map may include: determining one or more axes of symmetry from the arrangement of map tiles constituting the sample map; It may include randomly shuffling an object, and mirroring the object shuffled with respect to the first region to another region with respect to the axis of symmetry.

The learning step may include: determining a first degree of similarity of the similarity map by comparing the distribution of the game map characteristic data with the characteristic of the similarity map; comparing the characteristic of the sample map with the characteristic of the similarity map; determining a second degree of similarity of the characteristic of the map; and evaluating the similarity map according to the first degree of similarity and the second degree of similarity.

The learning step of the generating network may be ended when the first and second similarities reach a first threshold and a second threshold, respectively.

The generating of the map characteristic data distribution includes: classifying a game map into a plurality of characteristics including a reference characteristic and one or more auxiliary characteristics; generating data related to (hereinafter, “reference characteristic data”); setting reference coordinates of each of the plurality of reference properties in a three-dimensional space using the reference characteristic data of a plurality of game maps; For a game map, generating data on the auxiliary characteristic (hereinafter "auxiliary characteristic data"), converting the auxiliary characteristic data of a plurality of game maps into a three-dimensional vector, and using the three-dimensional vector as the reference It may include fitting based on coordinates, and determining similarity between game maps based on a relative distance between the fitted three-dimensional vectors.

The step of setting the reference coordinates of each of the reference attributes includes converting the reference attribute bit string into a reference attribute vector, a first eigenvector in which the variance of the reference attribute vector is maximized, and the first eigenvector and obtaining a plurality of orthogonal eigenvectors, including the first eigenvector, selecting three eigenvectors from among the plurality of eigenvectors in the order of increasing eigenvalues, and determining the first eigenvector group , setting the reference coordinates of each reference attribute in a three-dimensional space using the first eigenvector group as three axes.

The converting of the auxiliary characteristic data into a 3D vector may include converting the auxiliary characteristic data into an auxiliary characteristic vector, a second eigenvector in which the variance of the auxiliary characteristic vector is maximally preserved, and the second eigenvector and obtaining a plurality of orthogonal eigenvectors, including the second eigenvector, selecting three eigenvectors from among the plurality of eigenvectors in the order of increasing the size of the eigenvalue, and determining as a second eigenvector group , and projecting the auxiliary feature vector into a three-dimensional space using the second eigenvector group as three axes.

The fitting of the three-dimensional vector with reference to the reference coordinates includes the steps of matching directions of the first eigenvector group and the second eigenvector group and forming the auxiliary characteristic vector with respect to the reference coordinates. It may include placing an event.

According to the above configuration of the present invention, the time and cost required for designing a game map can be reduced by automatically producing a game map having characteristics similar to a sample map by using the sample map and the data normalized with the data related to the various characteristics of the real maps. can be reduced

1 is a block diagram illustrating a configuration of an apparatus for generating a game map according to an embodiment of the present invention.
2 is a flowchart illustrating a flow of a method for generating a game map according to an embodiment of the present invention.
3 is a block diagram showing the configuration of a computer game map similarity determination apparatus according to an embodiment of the present invention.
4 is a diagram schematically illustrating a format of reference characteristic data.
5 is a diagram schematically illustrating a process of generating mission characteristic data by a game map characteristic data generator according to an embodiment of the present invention.
6 is a diagram illustrating an example of classifying attributes of a game map attack direction according to an embodiment of the present invention.
7 is a diagram schematically illustrating a process of generating attack direction data by a game map characteristic data generating unit according to an embodiment of the present invention.
8 is a diagram illustrating an example of object arrangement data generation by the game map characteristic data generation unit according to an embodiment of the present invention.
9 is a diagram illustrating a result of mapping characteristic data of a plurality of maps into a 3D space by a 3D fitting unit according to an embodiment of the present invention.
10 is a flowchart illustrating a flow of a method for determining game map similarity according to an embodiment of the present invention.
11 is a flowchart illustrating a process performed by the reference coordinate setting unit in the reference coordinate setting step.
12 is a flowchart illustrating a process performed by the 3D fitting unit in the mapping step of auxiliary characteristic data into 3D space.

Since the present invention can have various changes and can have various forms, specific embodiments are illustrated in the drawings and described in detail in the text. However, this is not intended to limit the present invention to the specific disclosed form, it should be understood to include all modifications, equivalents and substitutes included in the spirit and scope of the present invention. In describing each figure, like reference numerals have been used for like elements. In the accompanying drawings, the dimensions of the structures may be exaggerated than in reality for clarity of the present invention.

Terms such as first, second, etc. may be used to describe various elements, but the elements should not be limited by the terms. The above terms are used only for the purpose of distinguishing one component from another. For example, without departing from the scope of the present invention, a first component may be referred to as a second component, and similarly, a second component may also be referred to as a first component.

The terms used in the present application are only used to describe specific embodiments, and are not intended to limit the present invention. The singular expression includes the plural expression unless the context clearly dictates otherwise. In the present application, terms such as “comprise” or “have” are intended to designate that a feature, number, step, operation, component, part, or combination thereof described in the specification is present, and includes one or more other features or It should be understood that the existence or addition of numbers, steps, operations, components, parts, or combinations thereof does not preclude the possibility of addition. In addition, A and B are 'connected' or 'coupled' means that A and B are connected or combined because other components C are included between A and B in addition to A and B being directly connected or bonded. that will include

Unless defined otherwise, all terms used herein, including technical and scientific terms, have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs. Terms such as those defined in commonly used dictionaries should be interpreted as having a meaning consistent with the meaning in the context of the related art, and should not be interpreted in an ideal or excessively formal meaning unless explicitly defined in the present application. does not Further, in the claims of the method invention, the respective steps may be interchanged in their order unless the respective steps are explicitly constrained to the order.

Hereinafter, with reference to the accompanying drawings, preferred embodiments of the present invention will be described in more detail.

The "computer game map" of the present invention means a map used in a computer game, and the "computer game" is a game executed on a computer such as a desktop or notebook PC, as well as a game called a console. It includes both console games executed on dedicated terminals and mobile games executed on smartphones or tablets.

According to one embodiment of the present invention, computer game maps are classified by a plurality of features. Herein, the term “characteristic” refers to a classification criterion determined by various elements constituting a computer game map. For example, the game map includes the type of mission that the user must perform in order to clear the map (hereinafter, “the type of mission”), the arrangement of map tiles constituting the map, the shape of the map, the arrangement of objects, etc. It can have characteristics and can be classified according to these characteristics.

Each characteristic may have a plurality of attributes. For example, the characteristic "mission type" includes a matching type that completes a mission by combining blocks of the same color or shape, a hitting type that completes the mission by hitting a gimmick, and a conditional hit that must satisfy specific conditions when hitting a gimmick It can have attributes such as a type, an exit type that is completed only when exiting through a specific exit, and a tile type in which the mission is in the form of tiles.

One map may have two or more attributes for one characteristic. For example, a certain game map may have both hitting type and exit type properties.

According to an embodiment of the present invention, the computer game map may be classified by a plurality of characteristics, and may be selected as one reference characteristic from among the plurality of characteristics. The reference characteristic is a characteristic used to set reference coordinates in a three-dimensional space, which will be described later. Characteristics other than the reference characteristics will be referred to as "auxiliary characteristics".

The attribute of the reference characteristic is referred to as "reference attribute".

According to an embodiment of the present invention, the game map is classified by ① the type of mission, ② the combination of the direction the user must move to clear the game map (hereinafter, “attack direction”), and ③ the arrangement of the object. can be That is, according to an embodiment of the present invention, as a criterion for determining the similarity of the game map, three characteristics such as a mission type, an attack direction, and an object arrangement are used.

According to an embodiment of the present invention, among the three characteristics, a mission type may be set as a reference characteristic.

The above-described embodiment is for illustration, and other characteristics other than the above three characteristics may be used, and the number of characteristics used may be less or more.

1 is a block diagram illustrating a configuration of an apparatus for generating a game map according to an embodiment of the present invention.

The game map generation apparatus includes a generation network 100 , a determination network 200 , and a game suitability determination unit 300 .

The generative network 100 and the discriminant network 200 constitute a generative adversarial network (GAN).

GAN is an artificial intelligence learning technology that trains G through a competition structure between the generative model (G) of the generative network and the discriminant model (D) of the discriminant network so that G generates data with characteristics very similar to the measured data. . In other words, through a competition process involving G and D, G is learned to capture the distribution of the measured data, and can generate plausible synthetic data that do not differ in characteristics from the measured data. Specifically, the generative model G and the discriminant model D compete by the following objective function and are trained to find the Nash equilibrium point.

(Equation 1)

Here, x is actually measured data, that is, real data, and G(z) is virtual data generated by the generation network. The first term on the right-hand side indicates the probability that the real data x is actually determined, and the second term indicates the actual estimated probability of arbitrary virtual data generated from the normal distribution of the actual data.

The discriminant model of the discriminant network is trained so that the objective function is maximized, and the generative model of the generative network is trained so that the objective function is minimized.

The generating network 100 according to an embodiment of the present invention is configured to generate a game map similar to the sample map, that is, a similar map, by receiving an arbitrary game map (a sample map) as an input.

More specifically, the generation network 100 generates a similarity map having characteristic data similar to the characteristic data of the sample map. In this case, symmetry information of the sample map may be used.

A map used in a puzzle-type game such as a block matching game is composed of tiles on which objects such as blocks are placed. It is one of the very important characteristics because the progress of the game varies depending on the arrangement of tiles and the arrangement of objects within the tiles.

The generating network 100 obtains symmetry information from the map tile arrangement of the sample map. That is, the generation network 100 finds the symmetry axis or symmetry point of the map tiles constituting the sample map. Even if the layout of the map tiles is not completely symmetrical, if the map tile has a shape that is as close to symmetrical as possible with respect to a specific axis or point, the corresponding axis or point can be set as the symmetry axis or point of symmetry.

Next, the generating network 100 is configured to generate a similar map by setting one of a plurality of regions divided based on a symmetry point as a reference region, and randomly shuffling objects on map tiles of the reference region.

In this case, the generating network 100 sets a symmetry axis or a symmetric point, or randomly shuffling an object is executed according to the parameter values of the generative model inside the generating network 100 .

Next, a learning process of the discrimination network 200 will be described.

The determination network 200 determines whether the similarity map generated by the generation network 100 is similar to (close to real) characteristics of the actual game maps using the distribution of characteristic data of a plurality of game maps that are actually used.

More specifically, the determination network 200 compares the distribution of the characteristic data set extracted from the plurality of real game maps with the characteristic data of the generated similarity map to determine the similarity of the similarity map generated by the generation network 100 .

Meanwhile, according to an embodiment of the present invention, the determination network 200 may be configured to further determine the similarity between the input sample map and the generated similarity map, in addition to determining the similarity between the distribution of the actual data set and the similarity map. .

That is, the discrimination network may be configured to compare the characteristic data of the sample map, which is the basic data for generating the similarity map, with the characteristic data of the generated similarity map to determine the degree of similarity between the two.

Accordingly, the game map generating apparatus according to the present invention can generate a game map having characteristics similar to both the general characteristics of the game map and the characteristics of the sample map.

Learning of the discrimination network 200 is performed so that the objective function is maximized. That is, the characteristic data of a plurality of real game maps and the characteristic data of the similar map generated in a state where the parameters of the generation model of the generation network 100 are fixed are input to the determination model to increase the objective function of the determination model. Learning of the discrimination network 200 is performed by adjusting the parameters.

The learning process of the generative network 100 is performed opposite to that of the discrimination network 200 . That is, the learning of the generative network 100 is performed in such a way that the parameters of the generative model of the generative network 100 are adjusted so that the objective function is minimized.

As described above, the generative network 100 and the discriminant network 200 are competitively learned by modifying the parameters of the generative model and the discriminant model. When the evaluation of the discriminant network 200 with respect to a predetermined threshold value is reached, it is completed.

Since the similar map generated by the generation network 100 is generated based on the distribution of the characteristic data of the game map and the characteristic data of the sample map, only the characteristics constituting the game map are similarly reproduced. Therefore, when the generated similar map is used in an actual game, problems such as impossible to clear the corresponding map or the arrangement of objects contradicting the game progress rules may occur. Therefore, it is necessary to determine whether the similar map is suitable for the actual game, apart from the evaluation of the similar map performed by the discrimination network 200 .

According to an embodiment of the present invention, the game map generating apparatus includes a game suitability determining unit 300 that determines whether the similar map generated by the generating network 100 is suitable for an actual game.

The game suitability determining unit 300 determines whether the similar map generated by the generating network 100 is designed according to a corresponding game rule. The game suitability of the similar map determined by the game suitability determination unit 300 is transmitted to the generating network 100, and when it is determined that the generated similar map is not suitable for the game, the generating network 100 generates the similar map Discard and create another similar map.

In the embodiment of Fig. 1, the game suitability determining unit 300 has been described as determining the game suitability before the similar map generated by the generating network is transmitted to the determining network 200, but the game suitability determining unit 300 is It may be configured to determine the game suitability of the similar map after the evaluation on the similar map is completed by the determination network 200 . In this case, the game suitability determination unit 300 may be disposed at the rear end of the determination network 200 .

On the other hand, when the learning of the generation network 100 and the discrimination network 200 of the GAN is finished, the generation network 100 is used to produce a game map similar to an arbitrary sample map.

A process in which the learned generation network 100 produces a similarity map from a sample map is performed in the same manner as a process in which the generation network 100 generates a similarity map from the sample map during learning.

The similar map produced by the generating network 100 is transmitted to the game suitability determining unit 300 to determine whether it is suitable for use in a game. In this case, the method in which the determination unit 300 determines the game suitability of the similar map is the same as the method in which the determination unit determines the game suitability of the similar map during learning.

The GAN of the map generating apparatus of the present invention uses a characteristic data set of a plurality of game maps for learning.

The map generating apparatus of the present invention may include a game map characteristic data generating unit 400 for generating a game map characteristic data set for GAN learning.

According to an embodiment of the present invention, the game map characteristic data generation unit 400 may be configured to map the characteristic data of the game map to a three-dimensional space.

The game map characteristic data generating unit 400 will be described later with reference to FIGS. 3 to 12 .

According to another embodiment of the present invention, the similarity determining unit may be configured to obtain a Hamming distance of the object initial arrangement data converted into the m×n bit string and further use it to determine the similarity between game maps. .

According to an embodiment of the present invention, the generation network 100, the determination network 200, the game suitability determination unit 300, and the map characteristic data distribution generation unit 400 constituting the GAN perform their respective functions. It may be configured hardware, software, or a combination of hardware and software. For example, the generation network 100, the determination network 200, the game suitability determination unit 300, and the map characteristic data distribution generation unit 400 constituting the GAN include commands necessary for performing each of the functions described above. It may be configured by a processor of a computer device that performs processing according to software and instructions. At this time, the software constituting a part of the generation network 100 , the determination network 200 , the game suitability determination unit 300 , and the map characteristic data distribution generation unit 400 constituting the computer GAN is installed in the computer device in the form of an application. can be

Hereinafter, a method for generating a game map according to the present invention will be described with reference to FIG. 2 .

2 is a flowchart illustrating a flow of a method for generating a game map according to an embodiment of the present invention.

The embodiment of FIG. 2 is performed using the game map generating apparatus shown in FIG. 1 .

First, a distribution of characteristic data of an actual game map used for GAN learning and similarity determination is generated (S1000).

Generation of the distribution of the characteristic data of the game map will be described later with reference to FIGS. 3 to 12 .

Next, the GAN generation network 100 and the discrimination network 200 are trained using the map characteristic data set distribution and the sample map characteristics (S2000).

The learning process of the generating network 100 and the discriminating network 200 of the GAN is performed in the manner described in the learning process of the map generating apparatus of FIG. 1 .

That is, the generation network 100 of the GAN generates a similarity map similar to an arbitrary sample map, the discrimination network 200 estimates the similarity of the similarity map from the map characteristic data distribution and the characteristics of the similarity map, and the discrimination network 200 ) and the generative network 100 are performed in such a way that the parameters of the discriminant model and the generative model are adjusted so that the objective function becomes the maximum and the minimum, respectively, using the gradient of the objective function calculated from the discriminant network 200, and the similarity map This is continued until the similarity between the characteristic of , and the distribution of the map characteristic data reaches a predetermined threshold.

Since the GAN learning process of the map generating apparatus of FIG. 1 has been described in detail above, a detailed description thereof will be omitted.

When the similarity between the characteristic of the similar map generated by the generating network 100 and the distribution of the map characteristic data reaches a predetermined threshold, the learning is terminated.

When the learning is finished, the learning network 100 is used for map production. That is, a similar map similar to the sample map is generated by inputting an arbitrary sample map to the generated network that has been trained (S3000).

The method in which the learning network 100 produces the similarity map is performed in the same way as generating the similarity map from the sample map during learning.

That is, the generating network 100 finds the symmetry axis or symmetry point of the map tiles constituting the sample map, sets one of a plurality of regions divided based on the symmetry axis or the symmetry point point as the reference area, A similar map is created by randomly shuffling objects.

Since the process in which the generating network 100 generates a similar map from the sample map has been described in detail in the operation of the map generating apparatus of FIG. 1 , a detailed description thereof will be omitted.

Next, it is determined whether the similar map generated by the generation network 100 is suitable for the actual game (S4000).

Since the determination regarding the actual game suitability of the similar map has been described in detail in the operation of the game suitability determining unit 300 of the map generating apparatus of FIG. 1 , a detailed description thereof will be omitted.

If it is determined that the generated similar map is suitable for the actual game, the generated map is provided to the corresponding game (S5000).

If it is determined that the generated similar map is not suitable for the actual game, the generating network 100 regenerates another similar map similar to the sample map.

The map generating apparatus and method of the present invention use the distribution of the game map characteristic data set for GAN learning and similarity determination.

According to an embodiment of the present invention, the game map characteristic data generation unit 400 may be configured to map the characteristic data of the game map to a three-dimensional space.

Hereinafter, the operation of the game map characteristic data generating unit 400 will be described in detail with reference to FIGS. 3 to 12 .

3 is a block diagram illustrating the configuration of the game map characteristic data generating unit 400 according to an embodiment of the present invention.

The game map characteristic data generation unit 400 according to an embodiment of the present invention includes a computer game map characteristic data generation unit 410 , a reference coordinate setting unit 420 , and a three-dimensional fitting unit 430 .

The map characteristic data generating unit 410 generates data regarding a plurality of reference attributes included in the reference characteristic (hereinafter "reference characteristic data") and data regarding the auxiliary characteristic (hereinafter "auxiliary characteristic data") .

Hereinafter, an example in which the map characteristic data generating unit 410 generates the reference characteristic data and the auxiliary characteristic data by using three characteristics such as a mission type, an attack direction, and an object arrangement will be described.

<Generation of reference characteristic data>

The game map characteristic data generating unit 410 generates reference characteristic data according to the attribute of the reference characteristic for each game map.

The reference characteristic data is a bit string having a length equal to the number of attributes (reference attributes) included in the reference characteristic, and each bit constituting the reference characteristic data corresponds to the presence or absence of one reference attribute.

4 is a diagram schematically illustrating a format of reference characteristic data.

In the example of FIG. 4 , the reference attribute is classified into N reference attributes (RA_1 to RA_N), and the reference characteristic data is composed of an N-bit bit string. Each bit of the bit string corresponds to one reference property, and is coded as 1 if the game map has the corresponding reference property, and 0 otherwise.

The game map characteristic data generating unit 410 generates reference characteristic data according to this format and attributes of each map.

One map may have multiple attributes at the same time, and in this case, a plurality of 1's should be included in the bit string of the mission characteristic data. Accordingly, the game map characteristic data generating unit 410 may use a so-called multi-hot encoding method when coding the reference characteristic data.

A mission that a user must perform in order to clear a game map used in a computer game is called a mission, and it is one of the very important characteristics because the design of the map varies depending on the type of mission.

According to an embodiment of the present invention, as a reference characteristic, the type of the aforementioned mission may be used.

In this embodiment, the types of missions are broadly classified into five types as shown in Table 1 below. In other words, the mission type has five properties.

property matching type hitting type conditional strike exit type tiled subcategory HyperClear
BlockClear
GoalScore EggClear
JellyparkCLear
BellClear
StoneClear
ChainmanClear DiscolorClear
ChainClear
BossClear
CardMatchClear CakeMove
CrowClear
GoingGoalClear JellyClear
DrawingClear
ThiefClear

Each attribute can be further classified into sub-attributes, and the sub-classifications listed in Table 1 are an example. The matching type has a mission to put together blocks of the same color or shape, the striking type has a mission to hit a specific gimmick, and the conditional strike type has a mission to hit a gimmick under specific conditions. The exit type has a mission to get out of a specific exit, and the tile type has a mission to complete a mission on a specific tile.

The classification of Table 1 is only an example, and the mission type may be classified into attributes different from those of Table 1, and other attributes other than the attributes of Table 1 may be added or some of the attributes of Table 1 may be excluded.

The game map characteristic data generating unit 410 generates mission characteristic data for each game map according to the above-described attribute of the mission type.

In this embodiment, since there are 5 attributes of the mission type, the mission characteristic data is a 5-bit bit string.

Hereinafter, the mission characteristic data generation process of the game map characteristic data generating unit 410 according to the present embodiment will be described with reference to FIG. 5 .

5 is a diagram schematically illustrating a mission characteristic data generation process of the game map characteristic data generating unit 410 according to the present embodiment.

3, the mission characteristic data according to this embodiment has a length of 5 bits, and from the least significant bit to the most significant bit, a tile type (T), an exit type (E), a conditional strike type (C), a striking type (H), has a format in which the presence or absence of matching (M) attributes is arranged in order.

When information on an arbitrary game map Map_n is input to the game map characteristic data generating unit 410, the game map characteristic data generating unit 410 extracts the mission type attribute and uses the format according to the attribute of each map according to the format. Create mission characteristic data.

In the example shown in FIG. 5 , the game map Map_n has attributes of a hitting type and an exit type. Accordingly, the game map characteristic data generating unit 410 generates "01010" as the mission characteristic data according to the above format.

In this embodiment, the mission characteristic data is reference characteristic data, and is used to set reference coordinates and origins of each reference attribute in a three-dimensional space to be described later.

<Secondary property data>

(1) Generation of attack direction data

The attack direction is the direction the user must attack in order to clear the game map, and is one of the important factors that determine the shape of the map and the characteristics of the game progress. For example, in the case of a matching map used in a block matching game, the block must be moved to match the color or shape, and the direction of the block the user must move to clear the map may be the attack direction. Similarly, in the case of an exit-type map, the user must find an exit to escape, and the direction for moving to the exit may be an attack direction.

6 is a diagram illustrating an example in which the attack direction is classified into 15 attributes.

In the example of FIG. 6, the direction that the user should attack in order to complete the mission is classified into 14 directions such as up, down, left, right, upper left, upper right, lower left, and lower right, and 15 properties such as when it is difficult to judge. .

The example of FIG. 6 is only an example, and it is natural that the attribute of the attack direction may be set differently.

The attack direction data is a bit string composed of as many bits as the number of attributes of the mission type, and each bit constituting the attack direction data corresponds to the presence or absence of one attribute.

The format of the attack direction data and the generation process of the attack direction data are the same as the mission characteristic data described above with reference to FIGS. 2 and 3 .

That is, if the attack direction is classified into M attributes, the attack direction data is composed of a bit string of M bits. Each bit of the bit string corresponds to one attack direction attribute, and is coded as 1 when the game map has the corresponding attribute and 0 when the game map does not have the corresponding attribute.

The game map characteristic data generating unit 410 generates attack direction data according to this format and the attributes of each map.

One map may have multiple attack direction attributes at the same time, and in this case, a plurality of 1's should be included in the bit string of the attack direction data. Accordingly, the game map characteristic data generating unit 410 may use a so-called multi-hot encoding method when coding the attack direction characteristic data.

In this embodiment, since there are 15 attributes of the attack direction, the mission characteristic data is a 15-bit bit string.

Hereinafter, a process of generating strategy direction data of the game map characteristic data generating unit 410 according to the present embodiment will be described with reference to FIG. 7 .

7 is a diagram schematically illustrating a process of generating attack direction data by the game map characteristic data generating unit 410 according to the present embodiment.

As shown in FIG. 7 , the attack direction data according to the present embodiment has a length of 15 bits, and each bit has a format in which the presence or absence of 15 types of attack direction attributes shown in FIG. 6 is sequentially arranged. Numbers (① to ⑮) indicated in the attack direction of FIG. 7 are symbols attached to the attack direction shown in FIG. 6 .

When information on the game map Map n is input to the game map characteristic data generating unit 410, the game map characteristic data generating unit 410 extracts the attack direction attribute and attacks according to the attribute of each map according to the format. Generate direction data.

In the example shown in FIG. 7 , the game map Map n has two-way, upper-left and upper-left diagonal unidirectional properties. Accordingly, the game map characteristic data generating unit 410 generates "000000011000000" as the attack direction data according to the format.

Meanwhile, although the attack direction is used as one of the auxiliary characteristics in the present embodiment, the attack direction may be used as the reference characteristic.

(2) Generation of object arrangement data

Various objects can be arranged on the game map, and it is one of the very important factors because the shape of the map and the progress of the game change according to the arrangement of the objects.

The game map characteristic data generating unit 410 divides the game map into regions of m×n and determines whether an object is placed in each region. The game map characteristic data generating unit 410 generates an mxn matrix by coding 1 if the object is placed in the divided area, otherwise 0. Also, the game map characteristic data generating unit 410 may flatten the generated m×n matrix into an m×n bit string. The process of converting the object arrangement data matrix into a bit string may be performed by arranging the rows of the matrix in order or arranging the columns in order.

Fig. 8 is a diagram showing an example of object arrangement data generation.

The example of FIG. 8 is an example of generating object arrangement data by dividing the game map into 9 areas of 3×3.

The game map characteristic data generating unit 410 divides the game map into 9 regions of 3×3 to determine whether an object is placed in each region, and codes 1 in the area where the object is placed and 0 in the area where the object is not placed. Create a 3x3 matrix. Then, the game map characteristic data generating unit 410 converts the 3x3 matrix into a 9-bit bit string. In the example of FIG. 8 , since the bit string is generated by arranging the rows of the matrix in a line, the object arrangement data generated by the game map characteristic data generating unit 410 is “010011001”.

Although it has been described above that the object arrangement matrix is first generated and then converted into a bit string in generating the object arrangement data, the game map characteristic data generating unit 410 omits the matrix generation process and directly generates the object arrangement bit string. It may be configured to create

(3) Generation of auxiliary characteristic data

According to the present embodiment, the game map characteristic data generating unit 410 generates auxiliary characteristic data by concatenating the attack direction data and the object arrangement data. For example, the game map characteristic data generating unit 410 sequentially concatenates the attack direction data "000000011000000" generated in the example of FIG. 7 and the object arrangement data "010011001" generated in the example of FIG. 6 to "000000011000000010011001" as an auxiliary characteristic create with data

The above process is performed for a plurality of game maps to obtain data necessary for similarity determination.

<Setting of reference coordinates in three-dimensional space>

Referring back to FIG. 3 , the reference coordinate setting unit 420 determines the reference coordinates of each of the plurality of reference attributes in a three-dimensional space using the reference characteristic data generated by the game map characteristic data generating unit 410 . set

The reference property data of the game map is data regarding the presence or absence of a plurality of reference properties, and which reference properties the plurality of maps have may be displayed in a coordinate space. That is, since each bit constituting the reference characteristic data may be used as a coordinate component, the reference characteristic data may be treated as a vector. Since the reference characteristic data of each game map is data regarding the presence or absence of N attributes, it may be treated as an N-dimensional vector. Therefore, when N is greater than 3, the distribution of these properties cannot be expressed in a three-dimensional space, so it is necessary to reduce the data to three dimensions. In this case, the reference coordinate setting unit 420 sets the coordinate axis of the three-dimensional space, the first eigenvector in which the variance of the properties of the reference characteristic is maximally preserved, and the N-1 pieces of the first eigenvector orthogonal to the first eigenvector. It can be configured to find an eigenvector. From the N coordinate axes obtained in this way, three eigenvectors having the largest eigenvalue size among the eigenvectors including the first eigenvector are selected and set as the coordinate axes of the three-dimensional space.

The setting of the coordinate axes as described above may be performed using Principal Component Analysis (PCA).

Further, the reference coordinate setting unit 420 can determine the reference characteristic distribution of each game map by projecting the reference characteristic data of the plurality of game maps on the three-dimensional coordinate axis, and from this distribution, set the reference coordinates of each attribute can In this case, the reference coordinate setting unit 420 may set the average value of the coordinates of maps having a certain reference attribute as the reference coordinates of the corresponding reference attribute. For example, the average value of the coordinate values in the three-dimensional space of the maps having the exit type mission attribute may be set as the reference coordinate of the exit type mission attribute.

Meanwhile, the 3D fitting unit 430 maps auxiliary characteristic data of each map into the above-described 3D space. The three-dimensional mapping of the auxiliary properties is performed in a manner similar to the setting of the reference coordinates of the reference properties described above.

The auxiliary property data of the game map is data regarding the presence or absence of a plurality of reference properties, and which auxiliary property properties the plurality of maps have may be displayed in a coordinate space. That is, since each bit constituting the auxiliary characteristic data can be used as a coordinate component, the auxiliary characteristic data can be treated as a vector. Since the auxiliary characteristic data of each game map is data regarding the presence or absence of M attributes, it may be treated as an M-dimensional vector. Therefore, when M is greater than 3, the distribution of these properties cannot be expressed in a three-dimensional space, so it is necessary to reduce the data to three dimensions. In this case, the three-dimensional fitting unit 430 sets the coordinate axis of the three-dimensional space, the second eigenvector in which the variance of the properties of the auxiliary characteristic is maximally preserved, and M-1 pieces of the second eigenvector orthogonal to the second eigenvector. It can be configured to find an eigenvector. Three eigenvectors having the largest eigenvalue size among the eigenvectors including the second eigenvector are selected from the obtained M coordinate axes and set as the coordinate axes of the three-dimensional space.

The setting of the coordinate axes as described above may be performed using Principal Component Analysis (PCA).

Subsequently, the 3D fitting unit 430 maps the auxiliary characteristic data to the 3D space by projecting the auxiliary characteristic vectors into a 3D space having three eigenvectors including the second eigenvectors as main axes.

Meanwhile, the 3D fitting unit 430 is configured to fit the auxiliary data mapped in the 3D space based on the aforementioned reference coordinates. That is, the 3D mapping unit 300 is configured to fit the auxiliary data mapped in the 3D space. The main axis of the dimensional space coincides with the main axis of the three-dimensional space where the reference coordinates are set, and the mapping of auxiliary data is rearranged around the reference coordinates.

Through the above process, the characteristic data of each map may be arranged in a three-dimensional space.

FIG. 9 shows a result of mapping characteristic data of a plurality of maps into a 3D space by the 3D fitting unit 430 .

The main axes x, y, and z of the three-dimensional space are determined by the above-described method, and the coordinates of the characteristic data of each map are the result of projecting the characteristic data on this main axis. The coordinates in the three-dimensional space are O, and reference coordinates are set for each reference property in this space. In the illustrated example, OH is the reference coordinates of maps having the hitting type attribute as the reference characteristic, and OE is the reference coordinate of the maps having the exit type attribute as the reference characteristic. The three-dimensional fitting unit 430 fits the auxiliary characteristic data of each map to the reference coordinates and displays them in this three-dimensional space.

According to an embodiment of the present invention, the map characteristic data distribution generating unit 400 may further include a similarity determining unit (not shown). The similarity determining unit determines the similarity between the game maps based on the relative distance between the 3D vectors fitted to the 3D space.

An example in which the similarity determination unit determines the similarity between a plurality of maps will be described with reference to FIG. 9 .

For example, the similarity determination unit calculates the relative distances d12, d13, and d23 of the maps M1, M2, and M3, and compares these distances. When the calculated distance is d23 < d13 < d12, the similarity determining unit determines that the maps M2 and M3 are most similar in the order of the relative distances.

According to another embodiment of the present invention, the similarity determining unit may be configured to obtain a Hamming distance of the object initial arrangement data converted into the m×n bit string and further use it to determine the similarity between game maps. .

According to an embodiment of the present invention, the computer game map characteristic data generating unit 410, the reference coordinate setting unit 420, the three-dimensional fitting unit 430 and the similarity determining unit include hardware, software, and hardware configured to perform respective functions; Or it may be a combination of hardware and software. For example, the data generating unit 100, the reference coordinate setting unit 420, the three-dimensional fitting unit 430, and the similarity determining unit perform processing according to the instructions and software including instructions necessary for performing each of the above-described functions. may be configured by the processor of the computer device. At this time, the software constituting a part of the computer game map characteristic data generating unit 410, the reference coordinate setting unit 420, the three-dimensional fitting unit 430 and the similarity determining unit 400 may be installed in the computer device in the form of an application. there is.

Hereinafter, a method of determining the similarity of a game map using the above-described computer game map similarity determining apparatus will be described with reference to FIGS. 8 to 10 .

8 is a flowchart illustrating a flow of a method for determining game map similarity according to an embodiment of the present invention.

Referring to FIG. 8 , first, a game map is classified into a plurality of characteristics including a reference characteristic and one or more auxiliary characteristics ( S1100 ). Since the classification method of the reference characteristic and the auxiliary characteristic of the game map has been described above, a detailed description thereof will be omitted.

Next, data related to a plurality of reference attributes included in the reference characteristic, that is, reference characteristic data is generated ( S1200 ). On the other hand, the game map characteristic data generating unit 410 generates, for each game map, data related to the auxiliary characteristic, that is, auxiliary characteristic data ( S1300 ). These steps ( S1200 , S1300 ) are performed by the game map characteristic data generating unit 410 , which has been described in detail in the operation of the data generating unit 100 , and thus a detailed description thereof will be omitted.

The reference coordinate setting unit 420 sets reference coordinates of each of the plurality of reference attributes in the three-dimensional space by using the reference characteristic data (S1400).

9 is a flowchart illustrating a process performed by the reference coordinate setting unit 420 in the reference coordinate setting step ( S1400 ).

Referring to FIG. 9 , the reference coordinate setting unit 420 converts the reference attribute data into a reference attribute vector (S1410), obtains a first eigenvector in which the variance of the reference attribute vector is maximally preserved (S1420), the first Among the eigenvectors orthogonal to the eigenvector, two of the largest eigenvalues and a three-dimensional space having the first eigenvector as main axes are determined (S1430), and reference coordinates of each reference property are set in the three-dimensional space. (S1440) The processing is sequentially performed.

Since the specific operation of the reference coordinate setting unit 420 has been described above, a detailed description thereof will be omitted.

Referring back to FIG. 8 , the computer game map similarity determining apparatus maps auxiliary characteristic data of a plurality of game maps to a three-dimensional space ( S1500 ). This step is performed by the three-dimensional fitting unit 430 .

10 is a flowchart illustrating a process performed by the 3D fitting unit 430 in the mapping step S1500 of the auxiliary characteristic data to the 3D space.

Referring to FIG. 10 , the three-dimensional fitting unit 430 transforms the auxiliary feature data into an auxiliary feature vector ( S1510 ), obtains a second eigenvector in which the variance of the auxiliary feature vector is maximally conserved ( S1520 ), and the first Among the eigenvectors orthogonal to the eigenvector, a three-dimensional space is determined using two and a second eigenvector having the largest eigenvalues as main axes (S1530), and an auxiliary feature vector is projected into the three-dimensional space. (S1540) The processes are sequentially performed.

Since the detailed operation of the 3D fitting unit 430 has been described above, a detailed description thereof will be omitted.

The game map characteristic data generated by the game map characteristic data generation unit 400 through the above process is used for learning and similarity determination of the generation network 100 and the determination network 200 of the GAN of FIG. 1 .

In the detailed description of the present invention described above, although it has been described with reference to preferred embodiments of the present invention, those skilled in the art or those having ordinary knowledge in the art will have the spirit of the present invention described in the claims to be described later. And it will be understood that various modifications and variations of the present invention can be made without departing from the technical scope.

100: generative network 200: discrimination network
300: game suitability determination unit 400: game map characteristic data generation unit

Claims (22)

A game map generating device that generates a game map with similar characteristics using a generative adversarial network (GAN), comprising:
a generative network trained to generate a similar map having characteristics similar to characteristics of the game map by inputting a sample game map;
a discriminative network trained to evaluate the similar map using a game map characteristic data set related to the characteristics of a plurality of game maps and the sample game map; and
a suitability determining unit for determining game suitability of the similar map generated by the generation network;
A game map generating device comprising a. According to claim 1, wherein the generation network,
determining one or more axes of symmetry from an arrangement of map tiles constituting the sample map;
randomly shuffling an object to be placed on a map tile with respect to a first area among a plurality of areas divided by the symmetry axis;
and mirror the object shuffled with respect to the first region to another region with respect to the axis of symmetry. The method of claim 2, wherein the discrimination network,
determining a first degree of similarity of the similar map by comparing the distribution of the game map characteristic data set and the characteristics of the similar map;
comparing the characteristic of the sample map with the characteristic of the similarity map to determine a second degree of similarity of the characteristic of the similarity map;
A game map generating apparatus for evaluating the similarity map according to the first and second similarities. 4. The method of claim 3,
The apparatus for generating a game map in which learning of the generating network is terminated when the first and second similarities reach a first threshold and a second threshold, respectively. The method of claim 4, wherein the game map characteristic data,
reference characteristic data pertaining to the reference characteristic of each map, and one or more auxiliary characteristic data pertaining to the auxiliary characteristic of each map;
The reference characteristic is divided into a plurality of reference attributes,
The apparatus for generating a game map, wherein the reference characteristic data and the auxiliary characteristic data are mapped in a three-dimensional space. According to claim 5, wherein the mapping to the three-dimensional space,
setting the reference coordinates of each of the plurality of reference properties in a three-dimensional space using the reference characteristic data of the game map,
and fitting the auxiliary characteristic data based on the reference coordinates. The method of claim 6, wherein setting the reference coordinates of each of the reference properties comprises:
obtaining a first eigenvector in which the variance of the reference attribute vector with respect to a plurality of game maps is maximally preserved, and a plurality of eigenvectors orthogonal to the first eigenvector;
including the first eigenvector, selecting three eigenvectors in the order of increasing eigenvalues from among the plurality of eigenvectors to determine the first eigenvector group;
By setting the reference coordinates of each reference attribute in a three-dimensional space using the first eigenvector group as three axes,
A game map generating device that is performed. 8. The method of claim 7, wherein fitting the auxiliary characteristic data to the reference coordinates comprises:
obtaining a second eigenvector in which the variance of the auxiliary feature vector is maximally preserved, and a plurality of eigenvectors orthogonal to the second eigenvector;
Including the second eigenvector, three eigenvectors are selected from among the plurality of eigenvectors in the order of increasing eigenvalues and determined as a second eigenvector group,
An auxiliary feature vector is projected into a three-dimensional space using the second eigenvector group as three axes,
By aligning the directions of the first eigenvector group and the second eigenvector group, and arranging the mapping of the auxiliary characteristic vector around the reference coordinate,
A game map generating device that is performed. 9. The method of claim 8,
The reference attribute is the type of mission performed to clear the computer game map,
The auxiliary characteristic data includes data on a combination of directions that the user must attack in order to clear the computer game map (hereinafter, “attack direction data”) and data on the initial arrangement of objects (hereinafter “object initial placement data”). ), a game map generating device comprising a. 10. The method of claim 9, wherein the type of the mission is N (N>3 is a natural number),
The reference characteristic data is an N-bit multi-hot encoded bit stream in which each mission is coded as 0 or 1. The game map generating apparatus. 11. The method of claim 10,
The object initial arrangement data is an m×n matrix in which 0 or 1 is coded to indicate whether an object is arranged on each map tile constituting the map.
A game map generation method that uses a generative adversarial network (GAN) to generate a game map with similar characteristics,
generating a distribution of characteristic data of an actual game map used for learning of the GAN and determining similarity;
Using a sample game map for learning as an input, a similar map similar to the distribution of characteristic data of the actual game map is generated, and distribution of characteristic data of the actual game map A game map characteristic data set and the sample game map regarding characteristics of the plurality of game maps training the generating network and the discriminating network of the GAN to evaluate the similarity map using
generating a similar map similar to an arbitrary sample map by using the learning network; and
determining whether the similar map generated by the generation network is suitable for an actual game;
A method of generating a game map comprising a.
13. The method of claim 12, wherein the learning step and the step of generating the similarity map comprises:
determining one or more axes of symmetry from an arrangement of map tiles constituting the sample map;
randomly shuffling an object to be placed on a map tile with respect to a first area among a plurality of areas divided by the symmetry axis; and
mirroring the object shuffled with respect to the first region to another region with respect to the axis of symmetry;
A method of generating a game map comprising a. The method of claim 13, wherein the learning step,
determining a first degree of similarity of the similarity map by comparing the game map characteristic data distribution with the characteristic of the similarity map;
determining a second degree of similarity of the characteristic of the similarity map by comparing the characteristic of the sample map with the characteristic of the similarity map; and,
evaluating the similarity map according to the first and second similarities;
A method of generating a game map comprising a. 15. The method of claim 14, wherein the learning step of the generation network is terminated when the first and second similarities reach a first threshold and a second threshold, respectively. The method of claim 15, wherein the generating of the map characteristic data distribution comprises:
classifying the game map into a plurality of characteristics including a reference characteristic and one or more auxiliary characteristics;
generating, for each game map, data on a plurality of reference attributes included in the reference characteristic (hereinafter, "reference characteristic data");
setting reference coordinates of each of the plurality of reference properties in a three-dimensional space by using the reference characteristic data of a plurality of game maps;
generating, for each game map, data on the auxiliary characteristic (hereinafter "auxiliary characteristic data");
converting the auxiliary characteristic data of a plurality of game maps into a three-dimensional vector;
fitting the three-dimensional vector based on the reference coordinates; and
determining similarity between game maps based on the relative distance between the fitted three-dimensional vectors;
A method of generating a game map comprising a.
The method of claim 16, wherein the setting of reference coordinates for each of the reference properties comprises:
converting the reference attribute bit string into a reference attribute vector;
obtaining a first eigenvector in which the variance of the reference attribute vector is maximally preserved and a plurality of eigenvectors orthogonal to the first eigenvector;
determining the first eigenvector group as a first eigenvector group by selecting three eigenvectors from among the plurality of eigenvectors in the order of increasing eigenvalues, including the first eigenvector;
setting reference coordinates of each reference attribute in a three-dimensional space using the first eigenvector group as three axes
A method of generating a game map comprising a. The method of claim 17, wherein the converting of the auxiliary characteristic data into a three-dimensional vector comprises:
converting the auxiliary feature data into an auxiliary feature vector;
obtaining a second eigenvector in which the variance of the auxiliary feature vector is maximally preserved, and a plurality of eigenvectors orthogonal to the second eigenvector;
determining as a second eigenvector group by selecting three eigenvectors in an order of increasing magnitudes of eigenvalues from among the plurality of eigenvectors, including the second eigenvector; and
Projecting an auxiliary feature vector into a three-dimensional space using the second eigenvector group as three axes
A method of generating a game map comprising a. The method of claim 18, wherein fitting the three-dimensional vector based on the reference coordinates comprises:
matching directions of the first eigenvector group and the second eigenvector group; and
disposing the map of the auxiliary feature vector around the reference coordinates
A method of generating a game map comprising a. 20. The method of claim 19,
The reference attribute is the type of mission performed to clear the computer game map,
The auxiliary characteristic data includes data on a combination of directions that the user must attack in order to clear the computer game map (hereinafter, “attack direction data”) and data on the initial arrangement of objects (hereinafter “object initial placement data”). ) to create a game map containing The method of claim 20, wherein the types of the mission are N (N>3 natural number),
The reference characteristic data is an N-bit multi-hot encoded bit stream in which each mission is coded as 0 or 1. The method for generating a game map. The method of claim 21 , wherein the object initial arrangement data is an m×n matrix in which 0 or 1 is coded to indicate whether an object is placed on each map tile constituting the map.

Images