KR20030072640A - Game system and method in which cyber clone acts - Google Patents

Abstract

The present invention is a game system performed by a game performance module based on a DB (database) of probability information about a player's input tendency in a computer game and a probability information about a transition state of a game state caused by input. As a module for implementing a cyber-clone to perform the game on behalf of the game owner (master) as a master propensity learning module for learning the game propensity of the owner; A game surrogate module that allows the cyber clone to play the game in a manner similar to that of the owner instead of the owner; A game capability acquisition module that enables to acquire game ability autonomously in a game performance process; The present invention relates to a game system including a game correspondence module that allows a game player to perform a game differently.

Description

Game system and method in which cyber clone acts}

The present invention relates to a game system and a game method performed by a game performance module based on a DB (database) of probability information about a player's input tendency in a computer game and a probability information about a transition state of a game state caused by an input. It is about. More specifically, the present invention is a module for implementing a cyber clone to perform a game on behalf of the game owner (owner) as a master propensity learning module for learning the game propensity of the owner; A game surrogate module that allows the cyber clone to play the game in a manner similar to that of the owner instead of the owner; A game capability acquisition module that enables to acquire game ability autonomously in a game performance process; The present invention relates to a game system including a game correspondence module that enables different gamers to perform different games.

Background Art With the development of computer programming technology and the emergence of various multimedia means and communication means such as the Internet, computer games have been advanced and widely spread. However, regardless of the type of game and its implementation method, the game is conventionally made between gamers A, B, and C connected through the Internet as shown in FIG. 1A. In other words, all conventional games are based on games between people. However, there is an additional game automation engine as an auxiliary means.

The inventor of the present invention allows a virtual game player (hereinafter referred to as "cyber clone") to replace a specific game player (hereinafter referred to as "master") to perform the game on behalf of the owner in cyberspace. We have developed a new game method that can create different interests and fun. The cyber clone according to the present invention learns the game style of the owner to perform the game according to the owner's game style, and can acquire the game ability by itself based on the game result, and can also play different games according to the game partner. As a result, they have substantially the same or the same game performance as the host player.

Figure 1a is a conceptual diagram of a conventional game method

Figure 1b is a conceptual diagram of the game system of the present invention

Figure 2a is an overall configuration of the present invention

Figure 2b is a detailed configuration of the cyber clone game performance process

Figure 2c is a detailed configuration of the information database of the cyber clone learning process

Figure 2d is a detailed configuration of the host-oriented learning process by the cyber clone

Figure 2e is a configuration diagram in the initial setting and updating process of the learning information databases

3a is an operation flowchart of a game proxy execution module

3b and 3c is a flow chart of the operation of the propensity learning module

3d is a flow chart of operation of the game ability acquisition module

4 is a definition table of a finite state machine.

5 is a state transition diagram

6 is an example of a state change according to the state configuration and state transition;

7 is an example of a range variable representation of a state configuration.

8 is a group state diagram of similar states.

9 is a state transition diagram according to the input

10 is an input and group state diagram

11 is a group input and group state diagram

12 is a state transition diagram according to unit time

Figure 13 is a state transition diagram with probability added

14 is a conceptual diagram of an intermediate state

15 is an exemplary diagram of a state transition path during a game.

16 is an illustration of a learning progress path

17 is a group state transition according to the input

<Overview of invention>

1B is a system diagram for explaining a concept of a game method according to the present invention. Instead of the players A, B, and C, the cyber clones A, B, and C are playing games over the Internet. In the cyber clone game of the present invention, the owner and the cyber clone are identified with the owner, and when the owner has to do other things in a hurry, the game may proceed through the clone or the owner takes over while the clone is playing the game. You can also proceed. Cyclone may play the game in the absence of the owner and the opponents (human gamers or their cyber clones) designated by the owner, or may participate in various game competitions by themselves. To this end, the cyber clone game system should be provided with various modules, including a game competition holding module, a real time broadcasting module, a module for playing, and an interface module with the owner. The hardware base of the implementation may be implemented in a server / client manner or through a connection between PCs without a server based on peer-to-peer technology. Or it may be implemented in a mixed manner of the two ways. This has the advantage of enabling a variety of operations while minimizing the load on the server.

Before explaining the structure and operation of a specific embodiment of the present invention, the theory of the game method of cyber clone, which can be said to be the core of the present invention, and the game propensity learning and game ability acquisition method of the owner will be described. In the game system according to the present invention, a game is performed by a game performance module which expresses the input tendency and the state transition tendency as a probability, and is mainly based on the probability information on the input tendency of the host player and the probability information on the change tendency of the game situation. Assuming that it is performed, firstly, 1) modeling of a general game system will be described, and 2) a method of providing a cyber clone model applicable to any kind of game will be explained. It will be described the master game propensity learning, autonomous game ability acquisition and game correspondence method of the cyber clone which is the core of the present invention.

Modeling of a General Game System

In computer science, a computer is a machine whose output is determined by changing the contents of various memory stored data according to a given processing rule. This is how programs inside a computer work. Computationally speaking, every computer program can be basically expressed as a finite-state automaton defined in FIG. The state at this time may be referred to as the contents of the various memories. The memory includes a hard disk, RAM, registers and the like.

Based on this, all computer game systems can also be represented primarily as a state transition model as shown in FIG. In FIG. 5, the state is represented by a circle, the left end is the start state, the right end is the final state, and the rest are all possible states during game play. In the finite state machine of FIG. 4, the final state is one, but here, it is composed of several for the sake of understanding. For example, two final states are possible, given that the game is split when more than one player plays the game.

Here, a state may be described as a data string of a specific point of time stored in a memory in order to express a game situation from a mechanical point of view, but more specifically, from a game point of view, a state corresponding to a resource, an environment, etc. given to gamers may be described. It is the value of a variable at a specific point in time.

6 shows an example of a data structure representing a state and contents of values in the data structure change according to a state transition. The fields constituting the state may be satisfied with respect to a region of certain values as shown in the example shown in FIG. 7.

In addition, the state does not necessarily have to be stopped as in the above examples, and if a same or similar memory data change is made every time for the same input, the series of data changes may be grouped and expressed as a single state.

Furthermore, as shown in FIG. 8, several states that can be regarded as similar in a particular aspect may be grouped and expressed as a group state. In the group state, the inputs can also be bundled. For example, in the state transition diagram in which the state transition occurs according to the input, as shown in FIG. 9, the states are bundled as shown in FIG. Accordingly, the inputs may be represented by a and b, d and e, respectively. This grouping of states has the advantage of simplifying the representation of complex systems.

For example, in the figure above, when the input that binds a and b is A1 and the input that binds d and e is A2, the player enters only A1 and A2 when entering the group state where the states are bundled. The transition can be handled by the computer itself.

Furthermore, when automating the state transition in the integrated state as shown in Fig. 11, the player only needs to input the input A1. When A1 is input and transitioned to the integrated state, the state transition inside the integrated state is automatically performed in a predetermined manner. These integrated internal state transitions may be the same each time or may be different each time.

The above state transition model can be effectively used to represent many computer games whose state transitions are determined by input. For example, a game such as poker or chess is a model in which the transition to the next state is determined by the player's input, and a game such as a street fighter is a model in which the state transition is determined by two players' input. The input to the game system at this time is a combination of inputs of the two players. These games do not have any state transitions while all participating players are not typing. However, games such as Tetris and StarCraft can cause state transitions even if the player doesn't type. In addition, since a game such as Street Fighter is not determined only by its input from the game party's point of view, a state transition may be caused by the opponent player even when the input is not performed. Games in such a case is difficult to express as a model in which the state transition occurs simply by the input as described above.

So let's look at the big difference between games like poker and games like Tetris. Games like Tetris and StarCraft show that the elements of time have a big impact on state transitions. For example, in the case of Tetris, the game may be automatically terminated when a game is started and a certain time passes and the end condition is satisfied even if the game player does not input. In the case of StarCraft, when a player activates a specific automated module through input, the automation module automatically performs a game action that affects the state transition independently of the input of the player until there is an input for deactivating the player. . Therefore, even when there is no input from the player, such as Tetris, a state transition for game play may occur. In the case of StarCraft, in particular, the player may activate a certain automation module, wait for the state transitions generated by the automation module to proceed to the desired state and then take the next input.

State transition occurs in principle because some event occurs that changes the state before it transitions. The most representative event is the player's input. However, in the paragraph above, state transitions can occur even without input from the player. Then, the state transition that occurs even without the player's input should be interpreted as being input from game constituents other than the player. The game constituent other than the player may be an opponent player, the game system itself, or an automation module. Regardless of the player's input in Tetris, new fragment items appear over time because the game system itself is providing input for such a state transition, and there is an automation module in StarCraft that allows workers to build weapons. Once activated by the player, the automation module continues to produce weapons regardless of the player's next input. Weapons are obviously one of the key components of a state, so the automation module is doing input to cause state transitions.

The present invention aims to provide a cyber clone model that can be applied to any kind of game. Therefore, in order to be able to express different games with the same model as described above, it is necessary to comprehensively expand the concept of input, state, and state transition in the state transition model. First of all, the state transition is not caused only by the input of the player but also by the change of time so that the game system has a separate time system so that the time of a certain interval is the unit time and from the boundary of the unit time to the next state. Let the transition occur. As in the example shown in Fig. 12, the input of the player is applied to the state transition at the boundary of the unit time closest to the input time, and if the inputs of the multiple players occur between the boundaries of the same unit time, they are considered to occur at the same time. do. Therefore, it is desirable to set the interval of unit time to a time short enough that gamers do not feel.

In the relationship between the input and the state transition, when the input occurs, the state transition is made, but this means when there is no actual player input at the boundary of every unit time as shown in FIG. 12 to reflect the state transition even when there is no input from the player. Is assumed to have occurred. In a game such as poker, when no actual input is given, the state transition at the boundary of unit time is a virtual input, and the state transition to it is like returning to itself. In fact, a change in time can also be regarded as a component of the input. This is because it is naturally accepted that a state transition occurs every unit time when time is included in the field constituting the state.

Now games like poker and games like Tetris can be represented as a single state transition model. But this model alone is still not enough to represent games like StarCraft. The state so far refers to the situation of the game at a particular point in time. However, in a game such as StarCraft, an automation module that automatically performs an action that affects the game state may be activated or deactivated according to a player's input in a specific state. In other words, state not only acts as a description of the situation, but also takes actions such as output based on input. This must be reflected in the model so that games such as StarCraft can be represented. That is, when an input is applied to a specific state, the model is extended so that a function for outputting the command to activate (execute) or disable (exit) the automation module is embedded in the corresponding automation module. This means that the concept of state is a data representation of the state of the game and includes the ability to trigger specific events. In this case, states have the same characteristics as objects in object-oriented programming (OOP) (of course, it is not necessary to match states to objects in the OOP to represent the automation module being included in the game system. For example, certain automation modules in the state have fields related to activation, and depending on what their values are, the game system needs to enable / disable the automation modules.)

The concept of input, state, and state transition has been redefined to model various games with various state transitions. In the future, the following description will be required for the present invention. However, the above concepts can be modeled in the same or similar way as before the transformation even if little modifications are made. For example, games can be modeled using variations of the concept, such as "state transitions may occur in terms of unit time, but may occur immediately if input is given." In this case, however, there is no significant difference between this model and the model. Therefore, the concepts established to explain the contents of the present invention can be variously modified in a direction expressing the description of the present invention differently.

<Physical clone input decision principle>

We have modeled the game system. The following describes how the cyber clone, which is the core of the present invention, can be implemented in such a game system.

In the game, it can be said that all game-related objects except themselves are integrated into one game system. This is analogous to the idea that the winning and losing of the game depends greatly on the action of the player, and the game system in this case cannot be explained by the deterministic state transition model as described above. This is because the state transition may come out differently each time even if the player inputs the same input in the same state. In other words, when the input of the opponent player constitutes the input of the game system together with his own input, it can be expressed as the above state transition. It is composed only of input, but internal game system affects the state transition, so even if the same input is performed in the same state, different state transitions can occur every time. In addition to humans, it may be the computer itself, or an automation module may affect the state transition).

This phenomenon can be clearly expressed by introducing a probability concept into the state transition model. In other words, it assigns the probability that a transition will occur for each transitionable state when a specific input comes in a specific state. For every state in the game, we set the probabilities for input and state transitions. These settings are allocated one set per opponent player when the opponent player can be recognized. This is because the game propensity by gamers varies widely, so that the probability of state transition for input is different for each player. The cyber clone, which is the core of the present invention, is implemented based on the state transition model to which this probability is added.

The cyber clone of the present invention follows the game inclination of the game player as it is and then develops the game ability in the state. Therefore, CyberClone must be able to grasp the owner's game propensity. In this case, we need to be aware of which input the owner would prefer in a particular state, given the uncertainty in the game's propensity for the clone. This can be expressed by adding the probability for each input occurrence to the state transition model with probability. Fig. 13 is an example of a state transition diagram in which two kinds of probabilities are included.

As shown in FIG. 13, when a state, a probability for each state transition, and an input probability for a specific game are given, the cyber clone is basically an input such as Equation 1 to be taken in a specific state. Is determined by Is The value of x that maximizes the value If any variable other than x exists in the combination of other variables and possible values of x The value of x in the combination that maximizes the value). Equation (1) shows which state transition the player prefers to select in a particular state. ) And the probability of occurrence of a state transition according to the opponent's input and the uncertainty of the game system. ). Of course, the opponent may be the computer itself.

In this way, factors in consideration of the owner's disposition ( ) And other factors influenced by the actions of opponents, such as opponents, that interfere with achieving that disposition ( ) Must be included in the implementation of the cyclone. Of course, the method of reflecting the two elements need not be limited to Equation (1). Two parameters included in Equation 1, the probability of the input to occur in a specific state ( ) And the probability of state transition for states transitionable from the state ( Are stored in the input tendency DB and the state transition tendency DB shown in Fig. 2A, respectively, and are used by the game proxy execution module to determine the appropriate input.

(Extension 1)

Equation 1 may be variously expanded for reasons such as an approach for implementing a cyber clone, for example, considering the effectiveness of granting the learning ability of the cyber clone in the future. As an example, the game system may provide the CyberClone with desirable state transitions in advance, allowing them to gradually play better games than their owners. The formula in this case is the same as Equation 2. Probability in the formula The probability that the system provides is assigned a high probability towards the transition path of states that are likely to win the game. These probabilities may be obtained and assigned by the game system developer by monitoring the game progress of those who are very good at the game, or may be pre-assigned appropriately by analyzing the game. Probability Parameters for Desirable State Transitions In the system in which is used, it is stored in the storm state transition trend DB shown in Fig. 2A and used together with the above two parameters to determine an appropriate input by the game surrogate module.

(Extension 2)

Another example of expansion is to quantify the degree of the player's advantage for each state and to include it in Equation 1 as an element that evaluates the value of the state transition based on this. The formula including this is equation (3). Probability for each state transition The state reached after the state transition is determined based on the difference between the states that can be reached after the state transition in terms of the game environment, that is, the advantage of the game player in terms of resources, such as weapons, food, soldiers, and the like, for game performance. The reason for taking into account the favorable difference between the reachable states is that the sum of the probabilities for transitioning from a particular state to the reachable states must be one. Once this probability is established, it is assumed that there is no change in composition for each state, so this probability is fixed without change.

Equation 3 may be regarded as a model of a form in which a human player checks a game environment through a monitor and takes an action accordingly when a human player performs a game. In other words, the human player expresses a model that senses a change in the state of the game with eyes and ears and takes the best input in the present state so that the next state is advantageous to him. However, Equation 3 does not accurately model the human game player. Human gamers sometimes intentionally transition to one of the next possible transitions. This may be the case, for example, to trick an opposing player. Ultimately, of course, the action is performed in anticipation of returning to a state that favors you. The probability associated with the high multiplier game state transition path mentioned in Equation 2 Is a value that adds value to the path itself and can be seen as inherent in the long-term utility value of the state transition (hereinafter, this probability is referred to as "long-term utility value"). On the other hand, the probability of Equation 3 Considers only the short-term utility value of state transitions (hereinafter, this probability is referred to as "short-term utility value"). Thus alone does not reflect the characteristics of the game, such as the concept of strategy and tactical retreat. Therefore, the expression is extended as shown in Equation 4 to reflect this.

The subscript n in Equation 4 represents how many transitions in the present state are considered. Preferably, transition to the end state can be considered. However, it is difficult for real human gamers to do this, and given the current computing power, it is not easy to consider future transitions in complex games. In particular, the more parameters there are in the learning of the clones described below, the more difficult the learning becomes. Therefore, in AI, the value of n is generally set to a value in a range in which a computer does not have a heavy load.

However, it is difficult to correctly model the master who is a human game player when the cyber clone is implemented by setting n to a small value in Equation (4). Human gamers do not consider all future states from the point of view of strategy and tactics, but play games with important future states in mind. Fortunately, the CyberClon model can be extended in that direction. To this end, we introduce the concept of intermediate states or groups of states or a mixture of these. Intermediate states are states that represent different phases of the game in the middle of all states in the game system. Equation 4 is extended by expressing the probability of transition between these intermediate states as probability. Intermediate state in equation (5) Is Is one of the strategically important intermediate states that can be reached. In the case of the state group, the formula is similar to Equation 5, and in the definition, all possible states in the game system are classified based on similarity, so that the group states are easily processed and included in the state transition model. The combined method of these two methods is to group similar states around the intermediate state into one group state. Similarly, the formula applied to the clone model is shown in Equation 5. In any case, information about them should be prepared in advance within the game system and provided when a cyber clone is created. 14 is a conceptual diagram representing a coupling scheme.

Parameters newly included in Equation 5 and In the system in which is used, the state evaluation information DB and the intermediate state evaluation information DB shown in FIG. 2A are respectively stored and used together with the remaining parameters to determine an appropriate input by the game surrogate performing module. parameter Is stored together in one of the two DBs.

Equation (5) mentioned the concept of the intermediate state only with respect to the short-term utility value, but can be replaced or extended by the long-term utility value. However, since the long-term utility value is a probability that it is a favorable state transition for victory regardless of the short-term utility value of the state transition, it is not significant to consider future state transitions and can only add computer load. Rather, the long-term utility value of a single state transition It seems desirable to include only.

In addition, Equation 5 expresses probable factors that make a decisive contribution to selecting the best input and constructs a formula by multiplying them. However, modifications may be made without departing from the spirit of the present invention in consideration of the utility and effectiveness of the implementation. For example, it is also possible to construct the formula by addition as shown in Equation 6 by discriminating the contribution of each component rather than expressing it as a product, and this may be more preferable. Coefficient in equation (6) When you build the game system, you can set up optimally through several experiments, or you can have the clones adjust directly through learning. Of course, formulas that apply multiplication and addition differently by element are also possible.

In addition, it is also possible to modify the equation to express the portion expressed as a probability, such as a score (score) rather than the probability. In particular, it may be more desirable to quantitatively indicate the short-term utility value that is gained or lost through its state transition.

<How to learn cyber clone>

(Learning principle)

So far, we have described the principle that the cyber clone in the game system performs a game under the assumption that various probability parameters are set in consideration of the owner's game tendency and opponent. The following describes a learning method for obtaining such parameters. The parameter acquisition process, or learning, can be done in two forms. It is mainly learning at the stage of grasping the owner's game tendency in the early stage and learning at game stage autonomously in the process of playing the game.

In general, artificial intelligence, which is the basis of learning, can be said to be realized through modeling the human brain. A typical model is Neural Network, which most closely models the human brain. The human brain conducts memory and thinking through the connections between a myriad of brain cells called neurons and learns by adjusting their connection strength. It is a principle that the connection strength of neurons is strengthened to be more accustomed to the praised or successful work, and the connection strength of the neurons is weakened on the contrary to the scolded or failed work. That is, the strength of the connection between the neurons related to the direction is strengthened, and the connection strength of the neurons related to the direction of retreat is weakened, and the learning is performed. The neural network learning method is the same, and the expression method is a state transition model with probability. Learning of the cyber clone of the present invention is implemented in a similar manner. The core probability parameters that can be adjusted through learning in this cyclone system are the input probability for a specific state and the state transition probability for possible transitions. In a particular state, the probability of a certain input and the probability of a specific state transition are enhanced or weakened, meaning that the probability of the other inputs and state transitions in the same state are slightly weakened or strengthened slightly.

(Learning method by target)

We have talked about the learning principle of cyber clone. Based on these principles, we will explain what the CyberClone needs to learn and how each learning works. For simplicity and clarity, we first need to look at the major components that make up the cyclone. A game system that can generate cyber clones must provide an environment in which clones can be created for each player. That is, master learning, clone only, and counterpart input propensity DBs and state transition trend DBs, state evaluation information DB, intermediate state evaluation information DB, excellent state transition trend DB, input time information DB, clone initialization module, and game proxy performance. Module, owner-oriented learning module, game ability acquisition module, and game correspondence module should be provided. The type of DBs provided is a kind of probability metrics.

CyberClone has excellent game performance ability to perform basic game when the main learning input tendency DB and the main learning state transition tendency DB are initialized and provided as the initial tendency DB and the state transition tendency initial DB shown in FIG. 2C. Given the state transition trend DB, better game ability is provided. (It is preferable that the initial DBs do not reflect the different game tendencies and show the probability distribution required for basic game performance.) As mentioned above, the knowledge of the clone As the DB, the propensity DB and the state transition DB are classified into master learning, clone-only, and counterparting, respectively. The reason for the master learning is to keep the game propensity from the master and track how the updated cyber clone's game propensity is changed from the game propensity. This is to be used as a comparative measure not to deviate greatly. In addition, the reason for distinguishing between clone-only and opponent-use is to acquire and use general game ability and game ability according to opponent separately. In other words, clone-only is initially initialized for master learning, and internal probability parameters are adjusted little by little with the game execution ability acquired in the course of performing all game opponents and games. In this case, the probability of the game is acquired in the process of playing the game with the opponent, and the internal probability parameters are adjusted more actively than in the case of clone learning. Therefore, the input propensity DB and the state transition propensity DB used by the actual cyber clones for game performance are counterparts, and are initialized with the contents of the clone-specific DBs at the initial confirmation of the opponent game player and updated during the game execution. As a result, the data structure of the master learning, clone-only, and opponent-adaptive DBs is the same, and only the content is changed through the game execution process.

The following three methods can be considered in learning the game propensity of a master who is a human game player. A configuration illustrating this is shown in FIG. 2D and will be described by specific embodiments. In other words, the master-oriented learning module is 1) a means of learning the owner's game propensity through a question-and-answer question and answer process with the master as a question-based learning, or 2) a direct game partner with the master as a direct peer learning (Dalian with the master). 3) one of means for learning the game propensity of the owner while playing the game, or 3) one of means for acquiring each probability value while monitoring the game of the owner as the host monitoring.

The questionnaire-based learning means of 1) prepares questions that can be appropriately used to grasp the owner's game propensity in advance and allows the questionnaire-based learning module to adjust the corresponding probability values according to the responses. This has the advantage of identifying the propensity of the owner in a short time, while the quality of the question becomes very important. The means for learning the game propensity of the owner by this questionnaire operates by receiving data from the questionnaire-based learning information DB. This DB provides multiple-choice questionnaires on how to react in the context of various gamers' inclinations in terms of strategy and tactics, and how to adjust probabilities related to propensity according to their answers. Is stored.

The direct training means of 2) is to adjust the probability values while playing the game with the owner in a game method performed by the direct training module which is pre-programmed so as to easily grasp the propensity of the owner. It is desirable that the game system be programmed to attack or defend the owner according to the state in which the inclination of the gamers occurs a lot in advance. The means of learning the game propensity of the owner by playing the game against the owner operates by receiving information from the Dalian Based Learning Information DB. This DB stores information about game states in which the player's disparity varies, and how to adjust the probabilities related to the propensity to the owner according to the reactions of the players in this state.

3) The host monitoring means causes the cyber clone to monitor the host gamer playing the game with various opponents, so that the host monitoring module shown in FIG. 2D monitors the relevant information from the Dalian based learning information DB as in the above method 2). It is to be provided to adjust the values in the master propagation input propensity DB and the master learning state transition trend DB.

The above three methods may be applied together as complementary relations to each other, rather than independently. Once the owner's disposition is identified, the clone will have the ability to play the game on its own, similar to the owner's disposition.

<Re-education of Cyber Clones>

The present invention also provides a system that enables real-time Internet relay of the game of the clone, but the game can be re-educated while the owner plays the game of the clone by allowing the game to be played after completion of the game. For example, the clone may be retrained in a manner of modifying its response by calling back the questionnaire-based learning module of the game propensity learning 1) of the owner (in this case, the query-based learning module may be a master learning input propensity DB and state). Readjust the probability values in the transition trend DB and apply them to the clone-specific input propensity DB and the state transition trend DB.)

Conversely, the game can be pointed out to the owner of the game's propensity to improve the game ability through repeated games. This is to inform the owner through a predetermined language when the probability value in the clone-only DB acquired while learning game ability through learning differs from the probability value in the master learning DB obtained from the owner. For example, when the probability obtained through a specific question / response among question / answers in the questionnaire-based probability adjustment information DB is significantly different from the probability obtained through learning, a method indicating that the response to the question is wrong. It is possible. For such a system that teaches the owner reversely, the master learning DBs containing the contents indicated by the initial owner should be preserved. By doing so, you can point out what's wrong with the owner by comparing the clones with updated internal values.

<Consideration of time factor>

The above explained the cyber clone that grows itself by acquiring the game propensity of the owner and repeating the game based on this. In the above description, for the sake of simplicity of explanation, the game learning of the clones was mainly focused on the input and the state transition, but one of the essential elements in learning the game propensity of the gamers is the time element. For example, some players have quick keyboard and mouse input, and vice versa. In this case, it is obvious that a person who has a fast input is advantageous to the game. Therefore, it is necessary to apply it to the state transition model. However, reflecting this directly on the formula has a problem that the formula is too complicated. Fortunately, the formula reflects some time. This is because the probability of a state transition in a state where the player makes a certain input and makes the next input fast and when it is not reached quickly and otherwise will be measured differently. Rather than reflecting the specific time factor in the formula, the time factor is reflected by measuring the waiting time for the next input of the player according to the state and input and installing and using a data structure that can reflect this for each state. Can be formulated, but since this makes the specification of the present invention difficult, the role of the time element as an element is described here as an example of a simple method.)

There are three types of considerations that a player may make with time delays in input. The first type is a case in which the player waits without input intentionally until the game enters a specific state by the reaction of an automation module or an opponent player. This case, however, is already implicit in the above-mentioned formula. The modeling of the game system has already been described, and it is assumed that even if there is no input at the boundary of the unit time where the state transition occurs, the state transition can be achieved. Under this premise, when a condition is satisfied and the actual player does not perform the input, the probability of occurrence () for the virtual input is relatively high, and the clone generates a state transition without input in such a state. will be.

The second type is when the player is limited in input speed by physical and physiological conditions. In this case, since it is limited by the force majeure input speed of the human player regardless of the conditions of the state, a separate processing system is required for this. For example, after entering a specific state through a certain input, the state transition is automatically performed every unit time without input, and a time device is installed in each state so that the clone cannot perform the next input until a certain time. That is, in each state, one of the various data structures describing the state sets a time mechanism, such as a latency variable. These variable values are all initially set to "0", and in the state where the variable value is reached by transitioning to the input generation in a specific state, the unit value is 1 unit in the transition state and the input waiting time of the master learned according to the input. Reset to minus time Then, when state transition occurs without input, the time reduced by one unit time is set as the variable value in the next state. The clone is forced to wait without performing input if the variable value is greater than zero under any condition. Since this reflects the force majeure of human beings regardless of the conditions of the game situation, it is preferable to process such simple processing rather than reflecting in the formula. Of course, because of the repetition of the game and the repetition of computer use, humans can improve the speed of input, so it is necessary to reflect this in the clone. This can be done by reducing the input wait time that has been learned based on the number and duration of the game by a small amount.

Finally, the third type is the case of delaying the input to take a short break during the game. In this case, it is not easy to actually include it in the clone game system. Because it's too uncertain what state the player wants to take a break from. However, it may be possible to group states into large groups to learn in which group states the host player tends to rest. Also, based on the fact that the dominant is likely to rest in a player's own favor, the clone can be rested at that point by detecting if the state continues to be very high across multiple state transition paths where the probability of virtual input is close. You will be able to take events. The processing method in the clone game system for this case is to generate a rest event when the state transition enters the group state and to perform input again when exiting the group state or to reach the state transition according to the state transitions. In the case of an advantage in the field, the input is performed again when the degree of deterioration is greatly deteriorated. Of course, in this case as well, the learning of the length of the owner's rest time allows the input to be interrupted within the range not exceeding the acquired length of the rest time.

Information related to the above is stored in the input time information DB shown in FIG. 2B and used in the input time determination module, which is a subordinate module in the game proxy execution module.

<Game automatic performance in the area where game disparity difference is small>

So far, we have described the game of Cyclone considering input, state transition, and time factors. We also described how to simplify the formula if necessary, taking into account the processing power of the computer. But the same is true for the whole process from start to finish. However, in the case of complex games, it may be necessary to seek further ease of implementation. To this end, it is also necessary to present a method that considers only state transitions in a small number of local state groups.

In general, when watching the game process of most gamers, there is a transition area of states that all take similar actions, and there is an area where transitions of states are varied by taking different actions according to game tendencies of each other. In this case, even if different actions are taken in the area where similar actions are taken, they do not affect game win or lose, or all gamers have a certain input pattern in this area. Therefore, in the case of clones, the master's game propensity is acquired in the whole process of the game rather than the game propensity in the whole process of the game, and the same effect as in the whole process can be obtained. In this case, the clone will automatically play the game in areas where similar actions are taken, which will already provide such an environment from the game system. The automatic game execution module shown in Fig. 2b plays this role.

Even in the case of a clone that learns the owner's propensity only in the partial regions as described above and proceeds the game accordingly, the start and end states can be set, and thus the above-described formula can be applied as well. Learning through the game process is also possible by adjusting the probability parameter values based on the termination state in the partial region, which can make a decision similar to the win / loss in the partial region. There may be a slight difference in the method of adjusting the probability parameter value for each partial region. Because the effect of the exit status of all partial areas on determining the final win or loss of the game will not be the same.

<Cleanup>

As described above, the model for the clone game system of the present invention may be variously established according to its needs. But basically, there are probability parameters in state transitions and input occurrences, which determine how the clones play the game and learn how to adjust the parameters. For example, consider a modified model that distinguishes time elements and inputs as shown in FIG. 17 again shows the concept of group state. Previously, however, the group state concept has been described primarily as a group of substates in the formula for finding the best input for the host gamers' propensity for the current state. However, in this context, in general, state transitions are based on inputs. In this regard, group states are regarded as bundles of autogenous state transitions that occur between the player's inputs and inputs. This allows transitions between group states to occur on the premise of an input. However, the group state is a bundle of states, and the internal state reached according to the state transition occurring within each hour in the game system results in the group state at that time. That is, the group state can be said to have the characteristic that the field values constituting the group state dynamically change. In addition, the player's input to the group state occurs when the time elapses due to the delay of the input time of the player or when the state changes suddenly according to the specific input of the opponent's game. do. As a result, it can be seen that the group state at the time of input generation in the deformation model is not very different from the specification so far as in the case of the input generation of the general state of the predeformation model mentioned above.

The above describes how to proceed with the game of the clone, how to learn the game propensity of the owner, how to grow the clone itself, and how to play the game individually to the opponent. The formulas presented so far are examples of implementing the invention and, of course, various modifications are possible under the same spirit.

The above specification describes the progress of every game as a transition of states and introduces probabilities where there is uncertainty. However, in AI theory, there are also ways of expressing problems with the same personality with rules and probabilities. When applying rules, apply them based on probabilities and select valuable rules through learning. However, only the expressions differ, but are basically similar to or the same as the state transition model. Therefore, of course, the present invention may be modified in various ways in the above description.

<Configuration and operation of the embodiment>

The present invention is a game system performed by a game performance module based on a DB (database) of probability information about a player's input tendency in a computer game and a probability information about a transition state of a game state caused by the input. It may be implemented as shown in FIG. 2A. That is, the game engine 10, the clone engine 20, can be composed of a knowledge database 30 of the clone.

The game engine 10 receives input information from the clone engine 20 and transmits the state information to the clone engine 20. The game state control module 11 and the game automatic progress module that control the state of the game (12), the game module interface 13 is composed of the constituent modules that the game engine should generally have in accordance with the characteristics of each game, and the clone interface module 14 for managing the interface with the clone engine 20 (game The engine is well known to those skilled in the art, and the specification of the game engine itself is omitted herein.

The clone engine 20 includes a clone initialization module 21 for setting initial values of various information DBs to be used by the cyber clone; A game proxy performing module 22 for performing a game on behalf of the owner; Master propensity learning module 23 for learning the game propensity of the master; A game ability acquisition module 24 for acquiring game capability by itself in a repeated game process; Consists of a game correspondence module 25 to perform a game differently for each game opponent, and receives information about the game status for each situation from the game engine 10 and provides various input information to the game engine (10). The clone engine 20 inquires information from the knowledge DBs 30 of the clones shown in FIG. 2A to play the game, and updates the contents of the relevant DB among the DBs with the learned contents.

The clone knowledge DB 30 includes: an input propensity DB 31 composed of probability information on input tendencies for each game situation of a master game player; A state transition trend DB 32 composed of probability information on a change tendency of the game situation; State evaluation information DB 33 composed of information for evaluating an advantageous degree of a game environment such as resources in each game situation; An intermediate state evaluation information DB 34 composed of information for evaluating the advantageous degree of the game in the intermediate state representing the states where the phase change is large in the game process; Superior state transition trend DB (35) consisting of information to guide the change of the desired situation for each game situation; It consists of an input time information DB 36 composed of information on an input speed, an input time, etc. of the host game player.

The core of the present invention lies in the game performance process of the cyber clone and the process of learning the cyber clone. And the game system associated with these processes is composed of more detailed elements.

Figure 2b shows a detailed configuration of the game execution process of the cyber clone. The game agent performing module for performing a game on behalf of the owner in the above 22, the input content determination module 221 to determine the appropriate input for each game situation by querying the information from the knowledge database 30 of the clone; An input time determining module 222 for determining an input time by querying trend information on the input time of the owner; It is composed of a detailed module such as a game automatic performance module 223 which automatically determines an input in a game situation in which the game propensity information of the owner is not inquired because the game propensity difference is hardly shown to the players.

The opponent correspondence input tendency DB 313 and the opponent correspondence state transition tendency DB 323 are appended DBs of the input tendency DB 31 and the state transition tendency DB 32 shown in FIG. In principle, each game partner is set up one by one and consists of information that enables the game to be played differently depending on the game partner. This is mentioned again in the learning process of the following cyber clones.

The operation of the game proxy execution module 22 configured as shown in FIG. 2B will be described with reference to the flowchart of FIG. 3A. First, the game proxy execution module 22 of the clone engine 20 obtains state information from the game engine 10 [1001], and displays the probability information on possible inputs and state transitions for this state. Inquiry is made from (30) [1003]. At this time, if the inquiry is not possible (for example, the game tendency difference is hardly shown to the players), the automatic determination is made by the automatic game execution module 223 [1005]. On the other hand, if the inquiry is possible, the input content determination module 221 applies the inquired information to the input determination formula to determine an input suitable for the state (that is, suitable for each game situation) [1007]. When the input is determined in this way, the input information is provided to the game engine 10 [1009], and state information of the next state is obtained from the game engine 10 [1011]. If the obtained state is the final state, the operation of the game proxy execution module 22 is terminated [1013]. If not, the process returns to step 1003 and repeats the same process [1015].

2C is a block diagram showing that the cyber clone in the clone engine 20 learns. Here, the input propensity DB 31 and the state transition propensity DB 32 have their associated DBs, respectively, the master learning input propensity DB 311 and the owner, which are composed of information acquired in the process of learning the owner's game propensity. Learning state transition trend DB 321; A clone-only input propensity DB 312 and a clone-only state transition trend DB 322 composed of information obtained in the process of acquiring game ability through repeated games of the cyber clone; It consists of the above-mentioned opponent application input propensity DB (313) composed of the information obtained in the process of acquiring the game performance ability for each game opponent and the opponent correspondence state transition trend DB (323).

The input tendency base DB 41 and the state transition tendency base DB 42 existing at the bottom of FIG. 2C store information capable of playing a game at an elementary level regardless of the game tendency of each player during a game system development process. The clones are copied to the master learning input tendency DB 311 and the master learning state transition tendency DB 321 by the clone initialization module 21 as DBs (see FIG. 2E).

Figure 2e is a schematic diagram showing the process of setting up and updating the learning information DB used in the present invention. Each of the master learning DBs 311 and 321 is an information DB reflecting the owner's game tendency, and is initialized by the clone initialization module 21 with the contents of the basic DBs 41 and 42, respectively. Is updated in the learning process. And the clone-only DB (312, 322) is an information DB reflecting the unique game ability of the cyber clone is initialized with the information in the master learning DB (311, 321) by the game proxy performance module 22 and the game ability acquisition module ( Is updated by 24). In addition, the opponent correspondence DBs (313, 323) is an information DB reflecting the ability to perform the game appropriately for each game opponent at the time of the actual game opponent and the clone-specific DB (by the game opponent correspondence module 25) ( The contents of 312 and 322 are initialized and updated by the game ability acquisition module 24, respectively. The distinction between clone-only DBs 312 and 322 and counterpart DBs 313 and 323 is that the former is a database of intrinsic game ability of the cyber clone, and it is a database in which the game ability can be applied to all game partners. In the latter, the initial contents are the same as the clone-only DB contents, but the game ability is learned and stored in accordance with the game opponent during the game opponent and Dalian. As a result, the DBs that the cyber clones inquire while playing games are the counterpart DBs 313 and 323.

On the other hand, there are additional components in the process of learning ownership. Figure 2d shows a detailed configuration of the process of the master propensity learning module 22 to learn the game propensity of the master. As shown in FIG. 2D, the owner propensity learning module 22 includes a questionnaire based learning module 22a for identifying a game propensity of the owner through a pre-produced questionnaire to grasp the owner's game propensity when designing a game; Or a direct-learning module 22b which grasps the owner's game propensity by acquiring the owner's response in a game situation showing different propensities for each player in the process of directly engaging with the owner; Alternatively, the owner monitors the process of playing the game with other gamers, and is implemented as host monitoring modules 22c which grasp the owner's game propensity, and these submodules are complementarily applied to the master propensity learning.

The questionnaire-based learning information DB 43 is a DB that the questionnaire-based learning module 22a inquires about, and information necessary to define the game propensity according to the surveys and the responses to the questionnaire prepared to easily grasp the inclination of the gamers. Consists of The Dalian-based learning information DB 44 is a DB that is inquired by the direct Dalian learning module 22b and the master monitoring module 22c. Therefore, it consists of information necessary to define the game propensity. The master tendency learning module 22 grasps the owner's game tendency with reference to the DBs 43 and 44 and reflects it in the master tendency input tendency DB 311 and the master learning state transition tendency DB 321.

The operation of the owner-oriented learning module 22 configured as shown in FIG. 2D will be described below with reference to the flowcharts of FIGS. 3B and 3C. First, among the methods such as question-based learning, direct peer learning, and master monitoring, a learning method is selected to determine which method to apply to master-oriented learning at a given time point [1101]. The algorithm when the question-based learning method or the direct peer learning method is selected is shown in FIG. 3B, and the algorithm when the master monitoring method is selected is shown in FIG. 3C. The algorithms are executed by the questionnaire-based learning module 22a, the direct peer learning module 22b, and the host monitoring module 22c, respectively.

1) When the questionnaire-based learning method is applied, the questionnaire and probability adjustment information is inquired from the questionnaire-based learning information DB 43 [1103]. If all the questionnaires have been processed and the inquiry is not possible, the procedure is terminated. [1111] If the inquiry is possible, the questionnaire is presented to the master game player and the owner's response is received [1105]. Thus, the process of adjusting the corresponding probability value in the master learning DB (311, 321) [1107] and retrieving the next questionnaire and the probability adjustment information [1109].

2) On the other hand, if it is checked in step 1101 that the direct peer learning method is applied, the game engine 10 is started [1113] and state information is obtained from the game engine 10 [1115]. Possible input information on the obtained state and probability adjustment information according to the owner's response for each input are retrieved from the Dalian-based learning information DB 44 [1117]. From the inquiry information, an input suitable for the above state is determined, the input information is provided to the game engine 10 [1119], and input information of the owner is obtained [1121]. Next, information of the next state is obtained from the game engine 10 [1123], and the corresponding probability value in the master learning DB 311 and 312 is adjusted based on the transition state and the owner's input and the probability adjustment information accordingly. [1125]. If the next state is the final state [1127], the master propensity learning process by the direct-learning module 22b ends, and if not, the process returns to step 1117 and the subsequent process is repeated.

3) On the other hand, if it is checked in step 1101 that the master monitoring learning method is applied, state information is obtained from the game engine 10 [1129], and input information of the owner is obtained [1131]. Then, the next state information is obtained from the game engine 10 [1133]. Based on the obtained information, probability adjustment information related to the input / state transition information is inquired from the Dalian-based learning information DB 44 [1135]. Check if the inquiry is possible [1136], if there is no information to proceed to step 1139, if there is information in the master learning DB (311, 321) based on the transition state and the owner's input information with the information inquired Adjust the probability value [1137]. If the next process is in the final state [1139], the owner-oriented learning process by the master monitoring learning module (2c) is terminated [1141], if not in the final state [1140] and returns to step 1129, the subsequent process is repeated. .

Now, the game capability acquisition process of the game capability acquisition module 24 will be described. The cyber clone of the present invention secures better game ability through self-learning than the ability given by the master by the game ability acquisition module 24 in the course of performing the repeated game after learning the propensity of the owner as shown in FIG. 2C. can do. This is based on the basic concept of artificial intelligence described above. When a clone wins as a result of a game transitioning through the values in a database, it increases the probability on the path and decreases it when it loses. The probability of adjustment is mainly an input occurrence probability and a state transition probability, but if the long-term utility value in the good state transition trend DB is used, it may be adjusted.) For example, as shown in FIG. As a result of the cyberclone performing the game as a path, when the final state indicates victory, as shown in FIG. 16, the probability values related to the path are adjusted upward to strengthen the connection between the states. On the other hand, when the final state reached in Fig. 15 means defeat, the various probability values existing in the reverse path shown in Fig. 16 are adjusted downward.

DB to be adjusted in the learning process is a clone-only input tendency DB 312, state transition trend DB 322, and relative correspondence input tendency DB 313 and state transition trend DB 323 shown in Figure 2c. As mentioned above, since the clone is dedicated to storing and using general game performance ability, the corresponding probability parameters are adjusted little by little according to the game result, whereas in the case of opponent application, the game performance ability for a specific opponent is adjusted. Because it is installed for storage and use, the probability parameters are adjusted more aggressively depending on the game result than for clone-only cases.

Self-learning through the game itself can be done through the adjustment of batch probability parameters after the game is completed. However, the learning can proceed as soon as the success or failure of local partial actions is confirmed during the game.

Now, the operation process of the game ability acquisition module 24 will be described with reference to FIG. 3D. First, all input and state transition path information performed during the game are acquired from the game proxy execution module 24 [1201]. Next, it is checked whether the game is a winning game [1203] or a losing game [1213]. In the case of a winning game, an input not reflected in the learning and the corresponding state transition are inquired on the state transition path [1205]. Check if the inquiry is possible [1207], and if the inquiry is possible, the input information in the clone-only input tendency DB 312 and the state transition tendency DB 322 and the state transition probability are adjusted to a predetermined value [1209]. . Next, the input information and the state transition probability in the input tendency DB 313 and the state transition tendency DB 323 for counterpart correspondence are adjusted upward to a predetermined value [1211]. On the other hand, if the inquiry is not possible in step 1207 [1206], the process ends [1221]. On the other hand, if it is checked as a game lost in steps 1203 and 1213, the state transition is inquired on the state transition path and the corresponding state transition [1215]. [1216] If the inquiry is possible, the input information and corresponding state transition probability within the clone-only input tendency DB 312 and state transition tendency DB 322 are adjusted to a predetermined value. The input information and corresponding state transition probabilities in the input tendency DB 313 and the state transition tendency DB 323 for relative correspondence are adjusted downward to a predetermined value [1219].

Lastly, when the opponent player is recognized by the game counterpart module 25 shown in FIG. 2C, the clone-only input propensity DB 312 and the state transition trend that are the products of the self-learning in the game process. This is accomplished by copying the DB 322 to the opponent application input propensity DB 313 and the state transition trend DB 323 and setting one per opponent player. Certain gamers cannot directly identify and learn opponent gamers, but grasp the game tendency of opponent gamers through indirect state transition. Therefore, as in the autonomous learning process, the game ability acquisition module 24 shown in FIG. 2C sets the opponent correspondence input tendency DB 313 and the opponent correspondence state transition tendency DB 323 which are set one by one for each opponent. By adjusting separately, it is possible to obtain an opponent propagation input tendency DB 313 and a state transition tendency DB 323 capable of appropriately performing a game for each opponent. By using the opponent-applied input tendency DB 313 and the state transition tendency DB 323 thus obtained, the cyber clone can perform a more effective game as the game is repeated with a specific opponent.

In the above, the configuration and operation of the specific embodiment of the present invention has been described with reference to the drawings, but not all of the elements shown in the drawings are necessarily used to implement the present invention, and according to the characteristics and development strategy of the game It will be apparent to those skilled in the art that components may be omitted or modified.

The above cyber clone can be applied to all existing computer games. There are various forms of games. Thus, slight modifications may be made for effective implementation but will not necessarily deviate significantly from the scope of the approaches described herein. Games such as StarCraft can be applied naturally because the formulas are made with this type of game in mind. In addition, games such as GoStop, Poker, etc. are characterized by the players taking action alternately so that the opponent's input does not directly affect the state transition of a particular player's input. The formula can still be applied since an estimate is possible based on their open paddles. However, in a game like Tetris, one player takes action alone and is heavily limited in time and discusses the performance of the game not by win or loss, but by game time and scores earned during the game. In this case, the game propensity learning for the master is the learning of typing speed and input preferences for each state (a habitual preference for the direction and number of rotations in the case of tetris), and the only factor that causes uncertainty in the state transition according to the input is meaning. May be weakened. However, in the case of Tetris where the floor moves or deforms in the middle of the input, as in the case of the modified Tetris, the input does not always have the same determinism in the next state. Still significant. In addition, the game ability improvement learning method in which a clone to play such a game grows on its own introduces the concept of victory as in other games when the player acquires more points for a longer time based on the master's average game completion time and score. If the game time is short and the acquisition score is low, the formula is still valid by introducing the concept of defeat as in other games. However, considering the fact that the game is decided in two factors, the length of time and the number of points obtained, it is possible to introduce a method of not learning when there is no winning or losing, or to win or lose if it is necessary. It is desirable to learn by doing.

The cyber clone game is expected to cause a big change in the game culture. The owner of the CyberClone may be studying at school or working at the company, but if you are curious about your CyberClone's game, you can check the progress of the game at any time by accessing the Internet. By playing games played by clones, you will be able to retrain your CyberClone or, conversely, acquire new skills from the CyberClone.

Claims (28)

Based on the probability information about the propensity of the owner's input and the tendency of the game situation to change according to the input, the master's game style is learned and the game is performed according to the owner's game style. It is a game system in which a cyber clone is operated, which is acquired and can play a different game for each game opponent through individual learning according to the game opponent. A game engine providing various modules and a game environment necessary for game execution; A clone initialization module for setting initial values of various information DBs to be used by the cyber clone; A game surrogate module for performing a game on behalf of the master; Host orientation learning module for learning the game orientation of the master; A game capability acquisition module that acquires game capability by itself in a repeated game process; A clone engine configured to play a game counterpart module for performing a game differently for each game partner, and receiving state information on a game state for each situation from the game engine and providing various input information to the game engine; And a knowledge database of clones that provide information to the clone engine and are updated with contents learned from the clone engine. The game system according to claim 1, wherein the knowledge database of the clone comprises an input propensity DB composed of probability information on input tendencies of game owners according to game status. The game system according to claim 2, wherein the knowledge database of the clone comprises a state transition trend DB composed of probability information on a change tendency of the game situation according to the input. The method according to claim 3, wherein the input propensity DB and the state transition trend DB, respectively, the master learning input propensity DB and state transition trend DB for learning and reflecting the game propensity of the owner; Clone-only input propensity DB and state transition trend DB to reflect game ability acquired in the clone's own learning process; It is set by the opponent gamers, including the input propensity DB and the state transition trend DB for the opponent to learn the game ability for each opponent game and store it so that the game can be performed differently for each opponent. A game system in which cyber clones are active. The method according to claim 4, the master learning input propensity DB and the state transition trend DB is initially set to the probability information required for basic game performance and is changed in the process of learning the owner's game propensity, clone-only input propensity DB and state transition The trend DB is initially set to values in the master learning input propensity DB and the state transition trend DB, and is changed in the process of acquiring game ability through repeated game play. The input propensity DB and the state transition propensity DB for opponent correspond to new game opponents. It is installed separately every time it encounters the initial value of the clone-only input tendency DB and the state transition tendency DB value is set in the cyber clone, characterized in that changed in the process of acquiring the game ability through the repeated game with the opponent game This game system is active. The method according to claim 3, wherein the knowledge DB of the clone is a state evaluation information DB consisting of information for evaluating the advantageous degree of the game environment, such as resources in each game situation; A game system in which a cyber clone is active, comprising: an intermediate state evaluation information DB composed of information evaluating a favorable degree of a game in an intermediate state representing a state in which a phase change is large in a game process. . The game system according to claim 6, wherein the knowledge database of the clones comprises an excellent state transition trend DB composed of information for guiding desirable situation changes for each game situation. The game system according to claim 7, wherein the knowledge database of the clone is composed of an input time information DB composed of information on an input speed, an input time, etc. of a master game player. The game system of claim 1, wherein the game proxy performing module of the clone engine comprises an input content determining module that searches for information in the clone's knowledge DBs and determines an appropriate input for each game situation. The game proxy performance module of claim 9, wherein the game engine performance module of the clone engine automatically determines an input in a game situation in which the game propensity information of the owner is not retrieved from the clone's knowledge DBs because the game propensity difference is hardly shown to the players. A game system in which a cyber clone is configured, further comprising a game automatic performance module. The game system according to claim 10, further comprising an input time determination module for determining the input time by querying trend information on the owner's input time from the clone knowledge DB. The method according to claim 1, wherein the propensity learning module of the clone engine Questionnaire-based learning module to grasp the owner's game propensity through a pre-made questionnaire to grasp the owner's game propensity when designing a game A questionnaire-based learning module DB for querying, a questionnaire-based learning information DB including questionnaires prepared to easily grasp a player's disposition and information necessary to define game disposition according to a response to the questionnaire; As a supplementary DB of the input tendency DB 31 and the state transition tendency DB 32, the master learning input tendency DB 311 and the master learning state transition tendency composed of information acquired in the process of learning the owner's game tendency, respectively. Cyber clone game system that is characterized in that the implementation is composed of a DB (321). The method according to claim 1, wherein the propensity learning module of the clone engine In the process of direct contact with the master, the direct-learning module (22b) which grasps the owner's game tendency by acquiring the owner's reaction in the game situation showing different tendency for each player, Dalian-based learning, which is a DB inquired by the direct Dalian learning module 22b, comprising information on a game situation in which a player's inclination is to be grasped and information required to define a game inclination according to a player's reaction in the situation. Information DB 44, As a supplementary DB of the input tendency DB 31 and the state transition tendency DB 32, the master learning input tendency DB 311 and the master learning state transition tendency composed of information obtained in the process of learning the owner's game tendency, respectively. Cyber clone game system that is characterized in that the implementation is composed of a DB (321). The method according to claim 1, wherein the propensity learning module of the clone engine The host monitoring module 22c that the cyber clone monitors the owner's game propensity while the owner plays the game with other players, As the DB inquired by the master monitoring module, the Dalian-based learning information DB (44) including information on the game situation in which the player's disposition is to be grasped and information necessary to define the game disposition according to the player's reaction in the situation. )Wow, As a supplementary DB of the input tendency DB 31 and the state transition tendency DB 32, the master tendency input tendency DB 311 and the master transition state transition tendency DB composed of information acquired in the process of learning the owner's game tendency, respectively. 321 is a game system in which the cyber clone is active, characterized in that it is configured and implemented. The method according to claim 1, wherein the game acquisition module 24 is As the clone executes the game, the input is determined through the values in the clone's knowledge database, and accordingly, the game is executed while the state transition occurs. Cyclone, characterized in that the up-probability values assigned to the input and state transition in the clone-only input propensity DB and clone-only state transition trend DB, and adjusts the values down when the game is lost This game system is active. The method according to claim 15, wherein the game acquisition module 24 is As the clone executes the game, the input is determined through the values in the clone's knowledge database, and accordingly, the game is executed while the state transition occurs. It is characterized by further adjusting the probability values assigned to the input and state transitions in the opponent-adaptive input tendency DB and the opponent-adaptive state transition tendency DB constituting the knowledge DB, and downgrading the values when the game is lost. Game system that includes cyber clone. The method according to claim 1, The game correspondence module 25, Recognizing a new opponent game player, the clone-specific input propensity DB 312 and the state transition trend DB 322, which are the products of the self-learning in the game process, are used for the new opponent response input propensity DB 313 and the state transition trend DB ( 323) by setting one per opponent gamers to allow the cyber clone to perform individual learning and game performance for each opponent game, the cyber clone is a game system. In claim 1, when the contents learned by the master tendency learning module and the contents acquired by the game ability acquisition module is different, the owner of the game by notifying the owner of the content to improve the game ability of the owner The game system in which the cyber clone is active, characterized in that it further comprises a master education interface module. Game engine that provides various modules and game environment necessary for game execution, Based on the probability information on the owner's input tendency and the probability information on the tendency of the game situation change, the cyber game player creates a cyber clone on behalf of the master game player, and the cyber clone learns the owner's game style. A clone engine that performs the game according to the game style and acquires the game ability by itself based on the game result, As a method of providing a game system consisting of a knowledge database of clones that provides information to the clone engine and is updated with contents learned from the clone engine, A clone initialization module that sets initial values of various information DBs to be used by the cyber clone in the clone engine, a game surrogate module that performs the game on behalf of the owner, a master propensity learning module that learns the game propensity of the owner, and repeated games Providing a game capability acquisition module for acquiring game capability by itself in a process, and a game correspondence module for performing a game differently for each game opponent; And providing a state transition tendency DB composed of probability information on a tendency of game situation change according to the input to the knowledge database of the clone. How to implement a game system that cyber clone is active. Game engine that provides various modules and game environment necessary for game execution, Based on the probability information on the owner's input tendency and the probability information on the tendency of the game situation change, the cyber game player creates a cyber clone on behalf of the master game player, and the cyber clone learns the owner's game style. A clone engine that performs the game according to the game style and acquires the game ability by itself based on the game result, As a method of providing a game system consisting of a knowledge database of clones that provides information to the clone engine and is updated with contents learned from the clone engine, A clone initialization module that sets initial values of various information DBs to be used by the cyber clone in the clone engine, a game surrogate module that performs the game on behalf of the owner, a master propensity learning module that learns the game propensity of the owner, and repeated games Providing a game capability acquisition module for acquiring game capability by itself in a process, and a game correspondence module for performing a game differently for each game opponent; An input propensity DB composed of probability information on input tendency of each game situation of the master game player in the clone knowledge database, a state transition tendency DB composed of probability information on a tendency of game situation change according to the input, resources in each game situation, etc. DB status information consisting of information evaluating the advantageous degree of the game environment of the game, intermediate status evaluation information consisting of information evaluating the advantageous degree of the game in the middle state that represents the state that changes a lot in the game process How to implement a game system in which the cyber clone is configured, including providing a DB. Game engine that provides various modules and game environment necessary for game execution, Based on the probability information on the owner's input tendency and the probability information on the tendency of the game situation change, the cyber game player creates a cyber clone on behalf of the master game player, and the cyber clone learns the owner's game style. A clone engine that performs the game according to the game style and acquires the game ability by itself based on the game result, As a method of providing a game system consisting of a knowledge database of clones that provides information to the clone engine and is updated with contents learned from the clone engine, A clone initialization module that sets initial values of various information DBs to be used by the cyber clone in the clone engine, a game surrogate module that performs the game on behalf of the owner, a master propensity learning module that learns the game propensity of the owner, and repeated games Providing a game capability acquisition module for acquiring game capability by itself in a process, and a game correspondence module for performing a game differently for each game opponent; An input propensity DB composed of probability information on input tendency of each game situation of the master game player in the clone knowledge database, a state transition tendency DB composed of probability information on a tendency of game situation change according to the input, resources in each game situation, etc. DB status information consisting of information evaluating the advantageous degree of the game environment of the game, intermediate status evaluation information consisting of information evaluating the advantageous degree of the game in the middle state that represents the state that changes a lot in the game process DB, excellent state transition trend DB consisting of information to guide the desired situation change by game situation, comprising the step of providing an input time information DB consisting of information about the input speed or input time of the host game, How to implement a game system with clones. The method according to any one of claims 19 to 21, wherein the game agent performing module 22 Obtaining state information from the game engine 10 [1001], and querying the clone knowledge DB 30 for probable information about possible inputs and state transitions for this state [1003], If the inquiry is not possible, the automatic determination is made by the automatic game execution module 223 [1005]. On the other hand, if the inquiry is possible, the input content determination module 221 inputs the inquired information to the input determination formula. Applying to determine an input suitable for the state (i.e., per game situation) [1007], If the input is determined, the input information is provided to the game engine 10 [1009], and the state information of the next state is obtained from the game engine 10 [1011], If the obtained state is the final state, the operation of the game proxy execution module 22 is terminated [1013], if not the final state, and returns to step 1003, and repeating the same process [1015] How to implement a game system with clones. The method according to any one of claims 19 to 21, wherein the owner-oriented learning module Querying the survey and probability adjustment information from the survey-based learning information DB (43), If all the questionnaire is processed and the inquiry is not possible, the procedure is terminated [1111]. If the inquiry is possible, the questionnaire is presented to the owner gamer and the owner's response is inputted. Adjusting the corresponding probability value in the master learning DB (311, 321) [1107], characterized in that learning the propensity to host by a question-based learning method comprising the step of querying the next questionnaire and probability adjustment information, How to implement a game system with clones. The method according to any one of claims 19 to 21, wherein the owner-oriented learning module Operating the game engine 10 [1113] to obtain status information from the game engine 10 [1115], Inquiring probable input information and probable adjustment information according to the owner's reaction for each input from the Dalian-based learning information DB 44 [1117], Determining input suitable for the state from the inquiry information and providing the input information to the game engine 10 [1119], and obtaining input information of the owner [1121], Obtaining information of the next state from the game engine 10 [1123], and adjusting the corresponding probability value in the master learning DB 311 and 312 based on the state transition and the input of the owner and the probability adjustment information accordingly. 1125], If the next state is the final state [[1127], if the owner-oriented learning process is terminated, if not the final state [1128] return to step 1117, the master process by direct direct learning method consisting of a step that repeats the subsequent process. The game system implementation method of the cyber clone, characterized in that learning. The method according to any one of claims 19 to 21, wherein the owner-oriented learning module Obtaining state information from the game engine 10 [1129], and obtaining input information of the owner [1131], Obtaining the next state information from the game engine 10 [1133], and querying probability adjustment information related to the input / state transition information from the Dalian based learning information DB 44 based on the obtained information [1135], Check if the inquiry is possible [1136], if there is no information to proceed to step 1139, if there is information in the master learning DB (311, 321) based on the transition state and the owner's input information with the information inquired Adjusting the probability value [1137], If the next process is in the final state [1139], the owner-oriented learning process by the master monitoring learning module (2c) is terminated [1141], if not in the final state [1140] to return to step 1129 the subsequent process is repeated A method of implementing a game system in which a cyber clone is active, characterized in that learning of owner propensity is performed by a master monitoring learning method including a step. The method according to any one of claims 19 to 21, wherein the game skills acquisition module 24 Obtaining all input and state transition path information performed in the game from the game proxy performing module 24 [1201], Checking whether the game is a winning game [1203] or a losing game [1213]; If the result of the check is a winning game, inquiring an input and a state transition not reflected in the learning on the state transition path [1205], Check whether the inquiry is possible [1207], and if possible, increase the probability values for the input and state transition in the clone-only input propensity DB 312 and state transition propensity DB 322 to predetermined values [1209]. Adjusting the probability values for the input and state transitions in the input tendency DB 313 and the state transition tendency DB 323 for a relative response to a predetermined value [1211], If it is checked as a game lost in steps 1203 and 1213, inquiring input and state transitions that are not reflected in the learning on the state transition path [1215], Check whether the inquiry is possible [1216], and if possible, lower the probability values for the input and state transition in the clone-only input direction DB 312 and the state transition trend DB 322 to a predetermined value [1217]; And adjusting the probability values of the input and state transitions in the input tendency DB 313 and the state transition tendency DB 323 for relative correspondence to a predetermined value [1219]. How to implement this game system. The game correspondence module 25 according to any one of claims 19 to 21, Copy the clone-specific input tendency DB 312 and the state transition tendency DB 322, which are the products of the self-learning in the game process, to the opponent corresponding input tendency DB 313 and the state transition tendency DB 323. A method of implementing a game system in which a cyber clone is performed, characterized in that the setting is performed one by one. The method according to any one of claims 19 to 21, wherein when the contents learned by the owner-oriented learning module differ from the contents acquired by the game ability acquisition module, the host gamer is notified about the contents. The game system implementation method of the cyber clone is characterized in that it can help improve the game ability of the owner.