LLM-based SLG game combat command method and system

Abstract

The invention provides an LLM-based SLG game combat command method and system, and relates to the field of data processing, and the method comprises the steps: receiving a dialogue of a command assignment intention initiated by a player in a window of a multi-user chat command room; the AI assistant determines a current player based on the background information of the player; the AI assistant determines a command type based on the command content of the current player; when the command type is an attack type, based on the operation information of the current player, environment data is obtained through an environment data type interface, and the AI assistant determines an attack strategy based on the dialogue content of the current player and the environment data and executes the attack strategy through a unit operation type interface; when the command type is a withdrawal type, environment type data is obtained based on the operation information of the current player, the AI assistant determines and executes a withdrawal strategy based on the dialogue content of the current player and the environment type data, and the method has the advantages that the flexibility, effectiveness and timeliness of combat command are improved through the artificial intelligence technology, and the game experience of the player is improved.

Description

Technical Field

[0001] The present invention relates to the field of data processing, and in particular to an SLG game battle command method and system based on LLM. Background Art

[0002] In the current gaming field, battle command usually relies on the player's immediate decision-making and operation. However, as the complexity of the game increases, the amount of information that players need to process increases dramatically, making it difficult for players to make quick and accurate decisions in battle. It is easy to make command errors, resulting in battle failure and poor gaming experience for players.

[0003] The existing game battle command methods rely on preset scripts and algorithms, which are limited in flexibility and adaptability. In addition, in terms of complex battle data, players need to rely on their own calculations and decisions to formulate operation strategies, which greatly increases the burden on players and causes player loss.

[0004] Therefore, it is necessary to provide an SLG game battle command method and system based on LLM, which is used to improve the flexibility, effectiveness and timeliness of battle command through artificial intelligence technology, and improve the player's gaming experience. Summary of the invention

[0005] The present invention provides an SLG game battle command method based on LLM, comprising: receiving a conversation of command assignment intention initiated by a player in a multi-person chat command room window; an AI assistant determines a current player based on background information of multiple players who initiate the conversation of command assignment intention, wherein the AI ​​assistant is established based on a large language model; the AI ​​assistant determines a command type based on the command content of the current player, wherein the command type is an attack type or a retreat type; when the command type is an attack type, based on the operation information of the current player, environmental data is obtained through an environmental data class interface, and the AI ​​assistant determines an attack strategy based on the conversation content of the current player and the environmental data, and executes the attack strategy through a unit operation class interface; when the command type is a retreat type, based on the operation information of the current player, environmental data is obtained through an environmental data class interface, and the AI ​​assistant determines a retreat strategy based on the conversation content of the current player and the environmental data, and executes the retreat strategy through a unit operation class interface.

[0006] Furthermore, the AI ​​assistant determines the current player based on the background information of multiple players who initiate the conversation with the command assignment intention, including: determining the current player based on the levels and intimacy of the multiple players.

[0007] Furthermore, the AI ​​assistant determines the intimacy of the player, including: obtaining the player's historical conversations; identifying topics and emotions corresponding to the historical conversations; and determining the player's intimacy based on the topics and emotions corresponding to the historical conversations.

[0008] Furthermore, the AI ​​assistant determines the command type based on the command content of the current player, including: when the conversation content of the current player includes coordinates, determining that the command type is an attack type; when the conversation content of the current player does not include coordinates, performing semantic analysis based on the conversation content of the current player to determine the command type.

[0009] Furthermore, based on the operation information of the current player, environmental data is obtained through an environmental data class interface, including: when the conversation content of the current player does not include coordinates, obtaining the current map of the current player, and obtaining environmental data based on the current map through an environmental data class interface; or, based on the map click information of the current player, obtaining target coordinates, and obtaining environmental data based on the target coordinates through an environmental data class interface.

[0010] Furthermore, the AI ​​assistant determines an attack strategy based on the conversation content of the current player and the environmental data, including: repeatedly generating new communication content based on the conversation content of the current player, analyzing the conversation content of the current player to the new communication content, and determining combat elements until all combat elements are collected; and generating an attack strategy based on all combat elements.

[0011] Furthermore, the AI ​​assistant determines a retreat strategy based on the conversation content of the current player and the environmental data, including: repeatedly generating new communication content based on the conversation content of the current player, analyzing the conversation content of the current player on the new communication content, and determining retreat factors until all retreat factors are collected; and generating a retreat strategy based on all retreat factors.

[0012] Furthermore, the method also includes: the AI ​​assistant generates report information on the battle process, battle results, retreat process or retreat results based on the character prompt words of the player assistant.

[0013] Furthermore, the method also includes: the AI ​​assistant updates the intimacy of the current player based on the conversation content of the current player.

[0014] The present invention provides an SLG game battle command system based on LLM, comprising: a dialogue receiving module, used for receiving a dialogue of command assignment intention initiated by a player in a multi-person chat command room window; a data interface module, including an environment data class interface and a unit operation class interface; an AI assistant, established based on a large language model, used for determining a current player based on background information of multiple players who initiate a dialogue of command assignment intention; the AI ​​assistant is also used for determining a command type based on the command content of the current player, wherein the command type is an attack type or a retreat type; when the command type is an attack type, the AI ​​assistant is also used for obtaining environment data through an environment data class interface based on operation information of the current player, the AI ​​assistant determines an attack strategy based on the command content of the current player and the environment data, and executes the attack strategy through a unit operation class interface; when the command type is a retreat type, the AI ​​assistant is also used for obtaining environment data through an environment data class interface based on operation information of the current player, the AI ​​assistant determines a retreat strategy based on the command content of the current player and the environment data, and executes the retreat strategy through a unit operation class interface.

[0015] Compared with the prior art, the SLG game battle command method and system based on LLM provided by the present invention has at least the following beneficial effects:

[0016] 1. Utilize the game map that the player focuses on while playing the game to automatically obtain the corresponding map battlefield space data, extract and model it, and express it in a digital way by regularizing the battle-related information data. Then, give it to the AI ​​assistant, so that it can perceive the battlefield space information based on the prompt words, and can communicate with the player only through coordinates or monster names, so as to achieve the effect of AI assistant perceiving battlefield information.

[0017] 2. Use the large language model to design AI assistants, including command analysis agents, combat analysis agents, retreat analysis agents, role-playing agents, and sentiment analysis agents. Through the collaboration between multiple agents, the full functionality of the AI ​​assistant is completed. Players can communicate with the AI ​​assistant in an anthropomorphic way through natural language and easily control various combat scenes in the game.

[0018] 3. By having normal conversations with AI assistants, players can get rid of the rigid problems of traditional games that rely on fixed commands and fixed processes, which greatly improves the player's in-game experience.

[0019] 4. Build multiple intelligent entities of the AI ​​assistant, combine the information revealed during the conversation with the battlefield information and combat rules information, and output complete command content that the program can recognize.

[0020] 5. Utilize the historical conversations between AI assistants and players, analyze the intimacy between AI assistants and players, and combine the player level to determine the service priority of game battle command. Through continuous, effective communication and personalized services, AI assistants can establish a deep relationship with players, thereby improving player loyalty and stickiness. Players will be more inclined to stay in the game and participate in more activities and challenges, which is crucial for the long-term operation of the game and community building. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] This specification will be further described in the form of exemplary embodiments, which will be described in detail by the accompanying drawings. These embodiments are not restrictive, and in these embodiments, the same number represents the same structure, wherein:

[0022] Figure 1 It is a flowchart of a SLG game battle command method based on LLM according to some embodiments of this specification;

[0023] Figure 2 It is a module diagram of an SLG game battle command system based on LLM according to some embodiments of this specification. DETAILED DESCRIPTION

[0024] In order to more clearly illustrate the technical solutions of the embodiments of this specification, the following is a brief introduction to the drawings required for the description of the embodiments. Obviously, the drawings described below are only some examples or embodiments of this specification. For ordinary technicians in this field, this specification can also be applied to other similar scenarios based on these drawings without creative work. Unless it is obvious from the language environment or otherwise explained, the same reference numerals in the figures represent the same structure or operation.

[0025] Figure 1 is a flowchart of a SLG game battle command method based on LLM according to some embodiments of this specification, such as Figure 1 As shown, a SLG (Simulation Game) game battle command method based on LLM (Large Language Model) may include the following process.

[0026] Step 110, receiving a dialogue of command assignment intention initiated by a player in a multi-person chat command room window.

[0027] The command assignment dialogue initiated by the player in the multiplayer chat command room window can include information such as enemy coordinates, enemy name, enemy location description, etc.

[0028] Just as an example, players can click on a certain location or an enemy fleet on the interface map, and a "Share coordinates to AI command channel" button will appear on the interface. Click this button to send the enemy coordinates to the command room chat channel. Players can also see the name of the monster on the interface, enter it in the chat window of the command room channel and tell the AI ​​assistant the name to attack, for example: Lisa, attack Katte's fleet. Among them, "Lisa" can be the role name of the player assistant played by the AI ​​assistant, and "katte" is the enemy name, which players can see and click on the interface. Players can have a normal conversation with the AI ​​assistant through the multiplayer chat command room window.

[0029] Step 120, the AI ​​assistant determines the current player based on the background information of multiple players who initiate the conversation with the command assignment intention, wherein the AI ​​assistant is established based on the large language model.

[0030] Specifically, the AI ​​assistant can be a collection of artificial intelligence functions that are coordinated by multiple agents, such as command analysis agent, combat assignment analysis agent, retreat analysis agent, role-playing agent, role-playing emotion analysis agent, etc. The agents here can be established based on a large language model.

[0031] In some embodiments, the AI ​​assistant determines the current player based on the context information of multiple players who initiated the conversation of the command assignment intention, including:

[0032] Based on the levels and intimacy of multiple players, the current player is determined. The higher the level, the lower the level. The player's intimacy can be queried from the intimacy list of all alliance players stored in the database. If the intimacy of the priority player cannot be found, the intimacy of the priority player is defaulted to 0. As an example only, the priority player with the highest intimacy can be used as the current player.

[0033] As an example only, put the player who is talking to the AI ​​assistant into a processing queue Q. Then judge the alliance rank of the player who issues the command later, and the ranks are divided from high to low: commander, general, colonel, and lieutenant. When there are players being served in the processing queue Q, the rank of the player who issues the command later is higher than the highest rank of the players in the queue Q, then his command will be processed first, and the conversation of the player who has been processed before will not be ended, that is, the AI ​​assistant will serve the player commands in the processing queue Q at the same time until the player commands are completed. If the rank of the player who issues the command later is lower than the highest rank of the players in the processing queue Q, his command will be rejected and the AI ​​assistant will reply, such as: "Other player commands are being processed, please wait". When the rank of the player who issues the command later is equal to the highest rank in the processing queue Q. Let the sentiment analysis agent analyze and output based on the intimacy of all players in the processing queue Q to determine the current player.

[0034] In some embodiments, the AI ​​assistant determines the player's intimacy, including:

[0035] Get the player's historical conversation;

[0036] Identify topics and emotions corresponding to historical conversations;

[0037] Determine the player's intimacy based on the topics and emotions corresponding to historical conversations.

[0038] As an example only, "'{content}"' is a conversation between a player and an AI assistant. The sentiment analysis agent can understand the context of the chat content, identify topics, analyze emotions, evaluate the atmosphere, and comprehensively evaluate the relationship between all characters. At the same time, the player's intimacy is re-evaluated. The player's intimacy evaluation should be expressed in percentage. The output format is as follows:

[0039] {

[0040] "Analysis":"The player gave attack orders to assistant Lisa",

[0041] "Emotion":"Half joking but a little serious",

[0042] "Evaluation":"80%",

[0043] "Conclusion":"The relationship between the player and Lisa is friends, with an intimacy of 80%"

[0044] }.

[0045] Specifically, the AI ​​assistant can calculate the number of historical conversations between the player and the AI ​​assistant in multiple historical periods (for example, one day, one week, one month, etc.), and calculate the first score based on the number of historical conversations between the player and the AI ​​assistant in multiple historical periods. The sentiment analysis agent calculates the quality parameters of each conversation based on the context of the chat content of each historical conversation, the identified topic, the analyzed emotion, and the evaluated atmosphere, and calculates the second score based on the quality parameters of each conversation. The first score and the second score are combined to calculate the player's intimacy.

[0046] Preferably, the AI ​​assistant can calculate intimacy based on the following formula:

[0047] I i =α 1 ×S (i,1) +α 2 ×S (i,2)

[0048]

[0049]

[0050] Among them, I i is the intimacy of the i-th player, α 1 and α 2 are weights, and α 1 and α 2 are greater than 0, α 1 +α 2 =1,β 1 and β 2 is the preset parameter, β 1 and β 2 Both are greater than 0, S (i,1) is the first score of the i-th player, S (i,2) is the second score of the i-th player, N (i,m) is the number of historical conversations between the ith player and the AI ​​assistant in the mth historical cycle, M is the total number of sampled historical cycles, Q (i,k) is the quality parameter of the kth historical dialogue of the ith player, and K is the total number of sampled historical dialogues.

[0051] Step 130, the AI ​​assistant determines the command type based on the command content of the current player.

[0052] The command type is an attack type or a retreat type.

[0053] In some embodiments, the AI ​​assistant determines the command type based on the command content of the current player, including:

[0054] When the current player's dialogue content includes coordinates, the command type is determined to be an attack type;

[0055] When the current player's dialogue content does not include coordinates, the command analysis agent performs semantic analysis based on the current player's dialogue content to determine the command type.

[0056] As an example only, retreat type commands include but are not limited to: 'recall the fleet', 'stop fighting', 'go home', 'come back', 'retreat', etc. Players may express their intention to retreat through concise colloquial language. Attack type commands include but are not limited to: the player sends an attack coordinate. The current player's dialogue content contains attack type commands and retreat type commands, and the type of the last command in the dialogue shall prevail.

[0057] The command analysis agent can output a json command type judgment result as follows:

[0058]

[0059]

[0060] Step 140, when the command type is an attack type, based on the operation information of the current player, the environment data is obtained through the environment data class interface. The AI ​​assistant determines the attack strategy based on the dialogue content and environment data of the current player and executes the attack strategy through the unit operation class interface.

[0061] In some embodiments, based on the operation information of the current player, obtaining environmental data through an environmental data interface includes:

[0062] When the current player's conversation content does not include coordinates, obtain the current player's current map, and obtain environmental data based on the current map through the environmental data class interface; or,

[0063] Based on the current player's map click information, the target coordinates are obtained, and the environmental data is obtained based on the target coordinates through the environmental data class interface.

[0064] Specifically, the environment data class interface is an HTTP interface used to obtain battlefield environment data. There are two ways to obtain environment data through the environment data class interface:

[0065] 1. The map where the player's perspective is focused is used as the basis for data acquisition, and the attributes of this map (for example, server attributes, galaxy attributes, etc.) are passed in. When the player sends non-coordinate content in the multiplayer chat command room window, the background monitoring program recognizes that the player does not send coordinates, and then uses the attributes of this map as parameters to call the environment data class interface corresponding to this map, obtain environmental data (for example, unit type data of all players or monster units in the map) and provide it to the AI ​​assistant, so that the player can feel that the AI ​​assistant has the ability to perceive the environment during the dialogue with the AI ​​assistant. For example, if the player only provides a target name, the AI ​​assistant can obtain and understand the attributes of the target name.

[0066] 2. Players click on a certain location on the map and share the map coordinates to the multiplayer chat command room. The background monitoring program recognizes that the player sends coordinates, and the AI ​​assistant obtains the unit type data of all players or monster units in the map that the player is concerned about through the attributes of the map where the coordinates are located. For example only, unit type data includes: player space station, player fleet, monster fleet, federal NPC fleet and other categories; information includes: name, shield, coordinates, enemy and friend attributes, mobile destination, etc. These data will be provided to the AI ​​assistant to allow it to perceive the description of the battlefield.

[0067] In some embodiments, the AI ​​assistant determines an attack strategy based on the current player's conversation content and environmental data, including:

[0068] Repeat the process of generating new communication content based on the current player's conversation content, analyzing the current player's conversation content on the new communication content, and determining combat elements until all combat elements are collected;

[0069] Generate attack strategies based on all combat elements.

[0070] Specifically, combat elements may include resources used in combat, combat objectives, etc. When combat elements are missing, the AI ​​assistant can automatically generate new communication content and complete the missing combat elements based on the current player's dialogue content on the new communication content. If the combat elements in the dialogue content are insufficient, the AI ​​assistant outputs the corresponding dialogue, communicates normally with the player or asks for more information, requiring the player to provide more content. When the player's dialogue contains content that is not related to the command, the AI ​​assistant can automatically generate a dialogue, requiring the player not to interfere with the normal release of commands.

[0071] For example only,

[0072] The latest fleet information we know is: {fleet_status}, the latest target information at the known coordinates is: {coordinates_info}, the attack strategy of the AI ​​assistant includes:

[0073] 1. The number of fleets to be dispatched, that is, the fleets to be dispatched in this command. Fleet ID is not required when talking to players. Please refer to the current fleet information: [{fleet_status}], where 'inst_id' is the fleet ID, 'Fleet_Status' is the fleet status, and 'Fleet_Name' is the fleet name. Please tell the player the number of fleets available in the conversation with the player, and communicate directly with the number of fleets. 'Fleet_Status' represents the current fleet status, which can be one of ['available', 'unavailable'], which is used to determine whether it can be used. Only fleets with a fleet status of "available" can execute commands. If there are no fleets available at present, please directly inform the player that the current fleet is busy and the retreat command must be used to recover the fleet first. If there are multiple fleets available, there is no need to specify the fleet ID. Instead, the AI ​​assistant will randomly match the fleets for task dispatch according to the number said by the player. The AI ​​assistant can decide the number of assignments by itself without asking, or confirm the number with the player. If there is only one fleet available, the AI ​​assistant has the right to dispatch the available fleet immediately without asking. If the player requests more fleets than the available number, please tell him clearly how many we have available. Just report to the player and explain the reason through Your_conversation in Json.

[0074] 2. The target, that is, a coordinate, requires the player to provide a coordinate, for example, (1,3760,-791,314), which is composed of four numbers. If the player has provided the coordinates, it is necessary for the AI ​​assistant to grasp the situation at this known coordinate. If there is a target near the coordinate, please give priority to the moving target, and then choose the closest target, but confirm with the player whether to attack it. If the player requires moving to the specified coordinates, abandon the nearby target and move to the specified position instead. All targets are one of ['enemy', 'neutral', 'friendly']. Both hostile and neutral targets can be attacked, and only friendly targets cannot be attacked. If it is a space station target, friendly (non-neutral) targets and targets with quantum interference fields are not allowed to attack. If the player insists on attacking, please ask the AI ​​assistant to reject the player's order and correct the player's mistake.

[0075] 3. Action type, which can be one of ["Attack-fleet", "Move", "Attack-monster"]. The AI ​​assistant must make its own judgment based on the known target information. If there is a target, it may be an attack, and if there is no target, it may be a move. When there are multiple target types, the player can be confirmed. "Attack-fleet" means that when there are player units such as player fleets or space stations near the coordinates, and the player's order is to attack, it means that the fleet will go to the specified coordinates to attack other player fleets or other players' space stations. "Move" means that when the distance of the targets near the coordinates is not 0, and the player requires to move to the specified coordinates and wait, it means that the fleet will move and dispatch without any attack instructions. "Attack-monster" means that when there are monster units near the coordinates, and the player's order is to attack, it means that the fleet will go to the specified coordinates to attack the corresponding scavenger monsters and other NPC monsters.

[0076] Step 150, when the command type is a retreat type, based on the current player's operation information, environmental data is obtained through the environmental data class interface. The AI ​​assistant determines the retreat strategy based on the current player's dialogue content and environmental data and executes the retreat strategy through the unit operation class interface.

[0077] In some embodiments, the AI ​​assistant determines a retreat strategy based on the current player's conversation content and environmental data, including:

[0078] Repeat the process of generating new communication content based on the current player's conversation content, analyzing the current player's conversation content on the new communication content, and determining the retreat elements until all retreat elements are collected;

[0079] Generate a retreat strategy based on all retreat factors.

[0080] Specifically, the retreat elements may include retreat resources, retreat locations, etc. As an example only, the retreat strategy may include assigned fleet information, including: fleet ID, fleet name.

[0081] The unit operation interface is divided into two interfaces: one is the attack interface, which is an http interface provided by the game background and is used to execute attack strategies.

[0082] As an example only, the interface parameters of the attack interface may be as shown in Table 1.

[0083] Table 1

[0084]

[0085] The attack interface functions include: commanding the AI ​​fleet to move, attack, garrison, etc. This function will form rules to allow the attack assignment analysis agent to output the corresponding fields and cooperate with engineering processing.

[0086] The unit operation interface can also include a retreat interface, which is an http interface provided by the game background and used to execute the retreat strategy.

[0087] As an example only, the interface parameters of the retreat interface may be as shown in Table 2.

[0088] Table 2

[0089]

[0090] The interface functions of the retreat interface include: letting the AI ​​fleet return to the home port and enter standby status.

[0091] In some embodiments, an LLM-based SLG game battle command method may also include step 160, where the AI ​​assistant generates report information on the battle process, battle results, retreat process or retreat results based on the character prompt words of the player assistant.

[0092] Specifically, in order to make the AI ​​assistant's reply more reflective of the character traits of the player assistant, the AI ​​assistant's role-playing agent is used, combined with historical dialogue interpretation, to report the battle process, battle results, retreat process or retreat results to the player commander in a more humane spoken dialogue according to the player assistant's personality prompts. The player assistant's personality can be manually selected by the player, or predicted by the AI ​​assistant based on historical dialogue.

[0093] For example, the report content before adjustment based on the player assistant's character prompts is:

[0094] Lisa:Commander DIABL0667, all three available fleets are dispatched to attack the space station'1nn1o0'at(15039,7737,-622,298).

[0095] The report content adjusted based on the player assistant's character prompts is:

[0096] Lisa:Commander DIABL0667, your three brave fleets are off to attack the poor space station'1nn1o0'at(15039,7737,-622,298).Do they even know what's coming?

[0097] In some embodiments, an LLM-based SLG game battle command method may also include step 170, where the AI ​​assistant updates the intimacy of the current player based on the conversation content of the current player.

[0098] Figure 2 is a schematic diagram of a module of an SLG game battle command system based on LLM according to some embodiments of this specification, such as Figure 2 As shown, an SLG game battle command system based on LLM may include a dialogue receiving module, a data interface module and an AI assistant.

[0099] The dialogue receiving module can be used to receive the dialogue of command assignment intention initiated by the player in the multiplayer chat command room window;

[0100] The data interface module may include an environment data class interface and a unit operation class interface;

[0101] The AI ​​assistant can be built based on a large language model to determine the current player based on the background information of multiple players initiating a conversation with the intent of command assignment;

[0102] The AI ​​assistant can also be used to determine the command type based on the command content of the current player, where the command type is an attack type or a retreat type;

[0103] When the command type is an attack type, the AI ​​assistant can also be used to obtain environmental data through the environmental data interface based on the current player's operation information. The AI ​​assistant determines the attack strategy based on the current player's command content and environmental data and executes the attack strategy through the unit operation interface;

[0104] When the command type is a retreat type, the AI ​​assistant can also be used to obtain environmental data through the environmental data interface based on the current player's operation information. The AI ​​assistant determines the retreat strategy based on the current player's command content and environmental data and executes the retreat strategy through the unit operation interface.

[0105] An SLG game battle command system based on LLM can be used to execute an SLG game battle command method based on LLM. For more descriptions of an SLG game battle command system based on LLM, please refer to the relevant description of an SLG game battle command method based on LLM, which will not be repeated here.

[0106] Finally, it should be understood that the embodiments described in this specification are only used to illustrate the principles of the embodiments of this specification. Other variations may also fall within the scope of this specification. Therefore, as an example and not a limitation, alternative configurations of the embodiments of this specification may be considered consistent with the teachings of this specification. Accordingly, the embodiments of this specification are not limited to the embodiments explicitly introduced and described in this specification.

Claims

1. A SLG game battle command method based on LLM, characterized in that: include: Receive the dialogue of command assignment intention initiated by players in the multiplayer chat command room window; The AI ​​assistant determines the current player based on the background information of multiple players who initiate the conversation with the command assignment intention, wherein the AI ​​assistant is built based on the large language model; The AI ​​assistant determines a command type based on the command content of the current player, wherein the command type is an attack type or a retreat type; When the command type is an attack type, based on the operation information of the current player, environmental data is obtained through an environmental data interface, and the AI ​​assistant determines an attack strategy based on the conversation content of the current player and the environmental data and executes the attack strategy through a unit operation interface; When the command type is a retreat type, based on the operation information of the current player, environmental data is obtained through an environmental data class interface, and the AI ​​assistant determines a retreat strategy based on the dialogue content of the current player and the environmental data and executes the retreat strategy through a unit operation class interface.

2. A SLG game battle command method based on LLM according to claim 1, characterized in that: The AI ​​assistant determines the current player based on the background information of multiple players who initiate the conversation of command assignment intention, including: The current player is determined based on the levels and affinities of the multiple players.

3. A SLG game battle command method based on LLM according to claim 1, characterized in that: The AI ​​assistant determines the player's intimacy, including: Get the player's historical conversations; Identify topics and emotions corresponding to the historical conversations; The intimacy of the players is determined based on the topics and emotions corresponding to the historical conversations.

4. A SLG game battle command method based on LLM according to claim 1, characterized in that: The AI ​​assistant determines the command type based on the command content of the current player, including: When the conversation content of the current player includes coordinates, determining that the command type is an attack type; When the conversation content of the current player does not include coordinates, a semantic analysis is performed based on the conversation content of the current player to determine the command type.

5. A SLG game battle command method based on LLM according to claim 1, characterized in that: Based on the operation information of the current player, environmental data is obtained through the environmental data interface, including: When the conversation content of the current player does not include coordinates, obtaining the current map of the current player, and obtaining environmental data based on the current map through an environmental data class interface; or, Based on the map click information of the current player, the target coordinates are obtained, and the environmental data is obtained based on the target coordinates through the environmental data interface.

6. A SLG game battle command method based on LLM according to any one of claims 1 to 5, characterized in that: The AI ​​assistant determines an attack strategy based on the conversation content of the current player and the environmental data, including: Repeat the steps of generating new communication content according to the conversation content of the current player, analyzing the conversation content of the current player with respect to the new communication content, and determining combat elements, until all combat elements are collected; Based on all the combat elements described above, an attack strategy is generated.

7. A SLG game battle command method based on LLM according to any one of claims 1 to 5, characterized in that: The AI ​​assistant determines a retreat strategy based on the conversation content of the current player and the environmental data, including: Repeating the steps of generating new communication content according to the conversation content of the current player, analyzing the conversation content of the current player with respect to the new communication content, and determining retreat elements, until all retreat elements are collected; Based on all the retreat factors, a retreat strategy is generated.

8. A SLG game battle command method based on LLM according to any one of claims 1 to 5, characterized in that: Also includes: The AI ​​assistant generates report information of the battle process, battle results, retreat process or retreat results based on the character prompt words of the player assistant.

9. The SLG game battle command method based on LLM according to claim 3, characterized in that: Also includes: The AI ​​assistant updates the intimacy of the current player based on the conversation content of the current player.

10. A SLG game battle command system based on LLM, characterized in that: include: A dialogue receiving module is used to receive dialogues of command assignment intentions initiated by players in the multiplayer chat command room window; Data interface module, including environment data interface and unit operation interface; An AI assistant, built on a large language model, is used to determine the current player based on the background information of multiple players in the conversation initiating the command assignment intention; The AI ​​assistant is further used to determine a command type based on the command content of the current player, wherein the command type is an attack type or a retreat type; When the command type is an attack type, the AI ​​assistant is further used to obtain environment data through an environment data interface based on the operation information of the current player, and the AI ​​assistant determines an attack strategy based on the command content of the current player and the environment data and executes the attack strategy through a unit operation interface; When the command type is a retreat type, the AI ​​assistant is also used to obtain environmental data through an environmental data class interface based on operation information of the current player. The AI ​​assistant determines a retreat strategy based on the command content of the current player and the environmental data and executes the retreat strategy through a unit operation class interface.

Images