Multi-agent collaborative interaction method, medium, computer equipment and system

Abstract

The invention relates to a multi-agent cooperative interaction method, a medium, computer equipment and a system, and the method comprises the steps: initializing a plurality of agents, obtaining a target or a sub-target of multi-agent cooperation, and obtaining an agent set based on the target or the sub-target; selecting a trigger agent, reversely deducing a plurality of interaction schemes, selecting an optimal scheme, and fusing the optimal scheme with the three-dimensional environment; multi-agent collaborative interaction is realized based on the optimal scheme; hardware is realized based on the method; the system comprises at least two agents configured in a VR equipment environment, and multi-agent collaborative interaction is realized by a controller for executing the method. According to the method, the selectable agents are obtained through different targets or sub-targets, cold start of collaborative interaction is achieved based on the trigger agents, a set of solutions is obtained based on the trigger agents, and then the optimal collaborative interaction scheme is selected; the method realizes multi-agent collaborative interaction, and is especially suitable for target execution work in a certain spatial range.

Description

Technical Field

[0001] The present invention relates to the technical field of control and regulation, and particularly relates to a multi-agent collaborative interaction method, medium, computer device and system. Background Art

[0002] An agent is an intelligent entity, which is an intelligent system built based on the cloud with AI at its core. The goal of an agent is the standard that the agent expects to achieve, which indicates the direction of the agent's activities. Each agent can have multiple executable goals, and the steps in the plan to complete the goals are completed by the actions that the agent can specifically execute. Of course, they can also be sub-goals.

[0003] As agents and their decisions are increasingly applied in specific scenarios, multi-agent collaboration is obviously a better choice. Compared with developing agents with multiple action types and complex action types, through multi-agent collaboration, specific operations (actions) can be shared among different agents, and only the collaboration between adjacent actions or sub-goals needs to be ensured.

[0004] In the prior art, the difficulty of multi-agent collaboration is mainly reflected in the fact that it often faces a non-static environment. This is because the behavior of each agent in the process will affect the environment, increasing the difficulty of cooperative learning. More importantly, in the scenario of multi-agent collaboration, there will also be a conflict between global and local optimality, and different agents may have different strategies and capabilities, which need to be coordinated and unified in the process of cooperative learning, which is more complex than single-agent learning. Summary of the Invention

[0005] To solve the above technical problems, the present invention provides a multi-agent collaborative interaction method, medium, computer device and system.

[0006] The technical solution adopted by the present invention is a multi-agent collaborative interaction method, which initializes multiple agents; obtains the goal or sub-goal of multi-agent collaboration, and obtains a set of agents based on the goal or sub-goal; selects a triggering agent, reversely deduces several interaction schemes, selects the optimal scheme and integrates it with the three-dimensional environment; realizes multi-agent collaborative interaction based on the optimal scheme.

[0007] Preferably, initializing the agent includes defining the movement range and executable action set of each agent in the current space.

[0008] Preferably, based on the movement range and movement set, the interaction characteristics of any two agents are obtained, and the interaction characteristics include the subspace block with overlapping actions and the docking envelope of the two agents in the interaction; If there is no interaction between any two agents, the interaction feature is an empty set.

[0009] Preferably, the actions are split layer by layer based on the goal or sub-goal until one or more sets of actions to be executed based on time series are obtained; and an agent set is obtained based on the set of actions to be executed and the spatial range involved in the goal.

[0010] Preferably, based on the goal or sub-goal, the agents are traversed according to the distance, and the agent closest to the goal or sub-goal that matches the action at the last time series split in each set of actions to be executed is selected, and this agent is used as the triggering agent corresponding to the current set of actions to be executed.

[0011] Preferably, after the triggering agent gives the initial action, in cooperation with its interaction feature, one or more actions for driving the previous agents are obtained in reverse, and the process is repeated until the derivation of each set of actions to be executed is completed, and the cost is calculated based on the docking envelope positions of every two agents to obtain the optimal decision.

[0012] Preferably, the interaction information between the docking envelope positions of two agents and the three-dimensional environment is obtained based on the optimal decision, and multi-agent collaboration is executed based on the interaction information.

[0013] A computer-readable storage medium stores a multi-agent collaborative interaction program thereon, and when the program is executed by a processor, the multi-agent collaborative interaction method described above is implemented.

[0014] A computer device includes a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the program, the multi-agent collaborative interaction method described above is implemented.

[0015] A multi-agent collaborative interaction system, the system includes: At least two agents, the agents are configured in a VR device environment; A controller is provided in cooperation with the two agents, and the controller executes the multi-agent collaborative interaction method described above.

[0016] The present invention relates to a multi-agent collaborative interaction method, medium, computer device and system. The method initializes multiple agents, obtains the goal or sub-goal of multi-agent collaboration, and obtains an agent set based on the goal or sub-goal; selects a triggering agent, reversely derives several interaction schemes, selects the optimal scheme and integrates it with the three-dimensional environment; realizes multi-agent collaborative interaction based on the optimal scheme; realizes the hardware based on the method; the system includes at least two agents configured in a VR device environment, and a controller that executes the method realizes multi-agent collaborative interaction.

[0017] The beneficial effects of the present invention are as follows: optional agents are obtained through different goals or sub-goals, and cold start of collaborative interaction is achieved based on the triggered agents. Based on this, a set of solutions is obtained from the triggered agents, and then the optimal collaborative interaction scheme is selected; multi-agent collaborative interaction is realized, which is particularly applicable to the execution of tasks by goals within a certain spatial range. Description of the Drawings

[0018] Figure 1 It is a flowchart of the method of the present invention. Detailed Embodiment

[0019] The following further describes the present invention in detail with reference to embodiments, but the protection scope of the present invention is not limited thereto.

[0020] The present invention relates to a multi-agent collaborative interaction method, which mainly includes the following steps: (1) Initialize multiple agents; (2) Obtain the goal or sub-goal of multi-agent collaboration, and obtain a set of agents based on the goal or sub-goal; (3) Select a triggered agent, reverse-derive several interaction schemes, select the optimal scheme and integrate it with the three-dimensional environment; (4) Implement multi-agent collaborative interaction based on the optimal scheme.

[0021] The following is an explanation based on specific steps.

[0022] (1) Initialize multiple agents; Initializing the agents includes defining the movement range and the set of executable actions of each agent in the current space.

[0023] Based on the movement range and the movement set, obtain the interaction characteristics of any two agents, where the interaction characteristics include the subspace block with overlapping actions and the docking envelope of the two agents during the interaction; If there is no interaction between any two agents, the interaction characteristics are an empty set.

[0024] In the present invention, in order to better realize multi-agent collaboration, the current space is divided during the initialization process. Based on the divided regions, the movement range of each agent is defined. Of course, during the implementation process, the executable actions of each agent are preset and unchanged; if the movement ranges of all agents are not sufficient to cover the current space, the uncovered regions in the current space are blocked, and unless the goal is located in this region, the size of the space to be planned is reduced; and if the goal is located in this region, the agent configuration is modified.

[0025] In the present invention, the interaction features of every two agents are subsequently established. In the subspace block where the actions of the two agents overlap, both agents can complete the actions (interactions). At the same time, the docking envelope of the two agents is set, generally an envelope sphere, so as to simplify the path planning, that is, in the path planning, it only needs to confirm that the docking envelope is in the overlapping subspace block, which greatly improves the application efficiency.

[0026] (2) Obtain the goal or sub-goal of multi-agent collaboration, and obtain the agent set based on the goal or sub-goal; Split the actions layer by layer based on the goal or sub-goal until one or more sets of to-be-executed actions based on time sequence are obtained; obtain the agent set based on the to-be-executed action set and the spatial range involved in the goal.

[0027] In the present invention, during the multi-agent collaboration process, the method can be directly executed based on the goal. However, in the actual application process, especially when the space is not large, generally, the subsequent steps are directly executed based on the goal, or a clear and unchanging sub-goal is selected. For example, if the goal is "make two cups of coffee in the current space and deliver them to the customer", then the sub-goals can be set as "deliver the coffee to the customer" and "make two cups of coffee"; Subsequently, split the actions layer by layer based on the goal or sub-goal until one or more sets of to-be-executed actions based on time sequence are obtained. For example, when executing the sub-goal of "deliver the coffee to the customer", it includes picking up the tray, placing the two cups of coffee on the tray one by one, picking up the tray, and delivering it to the customer; Finally, obtain the agent set based on the to-be-executed action set and the spatial range involved in the goal, such as the bar, the dining area, etc.

[0028] (3) Select the triggering agent, reverse-derive several interaction schemes, select the optimal scheme and integrate it with the three-dimensional environment; Based on the goal or sub-goal, traverse the agents according to the distance, and select the agent closest to the goal or sub-goal that matches the last to-be-executed action in the time sequence split in each to-be-executed action set as the triggering agent for the corresponding to-be-executed action set.

[0029] After the triggering agent gives the initial action, cooperate with its interaction features to reversely obtain one or more actions to drive the previous agents, and repeat until the derivation of each to-be-executed action set is completed. Calculate the cost based on the positions of the docking envelopes of every two agents to obtain the optimal decision.

[0030] In the present invention, the agent closest to the last to-be-executed action in the time sequence and available (able to execute this to-be-executed action) is used as the triggering agent. This triggering method is especially suitable for the cold start scenario, avoiding the situation that the training data cannot cover all training requirements in the case of diverse decisions; At this time, the agent is triggered to give an initial action. For example, if the action is to hand out coffee, then in line with its interaction characteristics, one or more actions that drive the previous agent are obtained in reverse. For example, if the triggered agent moves from the bar to the customer, then the action that the previous agent A needs to complete is to place the coffee in the food delivery area of the bar. And the previous agent B needs to complete the action of handing the made coffee to agent A after the coffee is made. Of course, in actual work, the actions may have a higher degree of continuity, such as the same agent completing the work of A and B mentioned above. After the derivation of each set of actions to be executed is completed, the cost is calculated based on the docking envelope positions of every two agents. The cost here includes, but is not limited to, the execution time and energy consumption of the paths of the two agents from the previous position point to the docking envelope position. Obtaining the optimal decision means the shortest execution time and / or the minimum energy consumption, which can be adjusted through weights.

[0031] (4) Implement multi-agent collaborative interaction based on the optimal solution; Based on the optimal decision, the interaction information between the docking envelope positions of the two agents and the three-dimensional environment is obtained, and multi-agent collaboration is executed based on the interaction information.

[0032] In the present invention, the interaction between the agent and the environment is realized through the see-through mode. In addition to the interaction between the agent and the user (real individual) that can be realized at each node, the real-time interaction information between the docking envelope position and the three-dimensional environment can also be obtained based on this. In this process, each agent can be applied to the VR environment to realize operations including but not limited to visualization and control the movement of other agents.

[0033] The present invention also relates to a computer-readable storage medium, on which a multi-agent collaborative interaction program is stored. When the program is executed by a processor, the multi-agent collaborative interaction method described above is realized.

[0034] The present invention also relates to a computer device, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the program, the multi-agent collaborative interaction method described above is realized.

[0035] The present invention also relates to a multi-agent collaborative interaction system, and the system includes: At least two agents, which are configured in the VR device environment; A controller is provided in cooperation with the two agents, and the controller executes the multi-agent collaborative interaction method described above.

[0036] The following gives an embodiment of the method of the present invention in the system.

[0037] Define the goal as making a cup of coffee in the current space and delivering it to the customer's hand; Initialize multiple agents, define the movement range and the set of executable actions of each agent in the current space, and finally obtain: Agent a: {the whole venue, cleaning}; Agent b: {the back kitchen, cooking}; Agent c: {the bar & the back kitchen, preparing drinks & serving meals (from the back kitchen)}; Agent d: {the bar & the dining area, cashier & cleaning}; Agent e: {the dining area, delivering food}; Agent f: {the bar, preparing drinks}; Agent g: {the office, preparing bills} The goal of the multi-agent collaboration is to make a cup of coffee and deliver it to the customer. Based on this goal, first exclude a, b, d, and g, and obtain the agent set {c, e, f}. Among them, the interaction features of any c and f, e and f, and e and d can be constructed, including the subspace blocks with overlapping actions and the docking envelopes of the two agents in the interaction; Split the actions layer by layer based on the goal or sub-goal. Then the current goal requires performing drink preparation and food delivery, and an interaction needs to be realized at the intersection of the bar and the dining area; Considering that e is for food delivery, a periodic agent can be selected as the trigger agent, and 2 interaction schemes can be derived by reverse deduction: ① Agent c prepares drinks → Agent e delivers food; ② Agent f prepares drinks → Agent e delivers food; Considering that the action of Agent f has a higher degree of simplicity, in the idle state, select the optimal scheme ② and integrate it with the three-dimensional environment, that is, obtain the interaction information between the position of the docking envelope of the two agents and the three-dimensional environment; Implement multi-agent collaborative interaction based on the optimal scheme.

[0038] Those skilled in the art should understand that the embodiments of the present invention can be provided as a method, a system, or a computer program product. Therefore, the present invention can adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present invention can adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0039] The present invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It should be understood that each flow and / or block in the flowchart illustrations and / or block diagrams, and combinations of flows and / or blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions executed by the processor of the computer or other programmable data processing apparatus create means for implementing the functions specified in the flowchart flow or flows and / or block or blocks. Figure 1 one flow or more flows and / or blocks Figure 1 or one block or more blocks.

[0040] These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means that implement the functions specified in the flowchart flow or flows and / or block or blocks. Figure 1 one flow or more flows and / or blocks Figure 1 or one block or more blocks.

[0041] These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process, such that the instructions executed on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and / or block or blocks. Figure 1 one flow or more flows and / or blocks Figure 1 or one block or more blocks.

[0042] Although the preferred embodiments of the present invention have been described, additional changes and modifications can be made by those skilled in the art once they learn of the basic inventive concept. Therefore, the appended claims are intended to be construed to include the preferred embodiments and all changes and modifications that fall within the scope of the present invention.

[0043] It is apparent that those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalent technologies, the present invention is also intended to include these modifications and variations.

Claims

1. A multi-agent collaborative interaction method, characterized in that: The method initializes a plurality of agents; Obtaining a goal or sub-goal of multi-agent collaboration, and obtaining an agent set based on the goal or sub-goal; Select the triggering agent, reversely deduce several interaction schemes, select the best one and integrate it with the three-dimensional environment; Realize multi-agent collaborative interaction based on the optimal solution.

2. A multi-agent collaborative interaction method according to claim 1, characterized in that: Initializing the agent includes defining the motion range and executable action set of each agent in the current space.

3. A multi-agent collaborative interaction method according to claim 2, characterized in that: Based on the motion range and motion set, the interaction features of any two agents are obtained, wherein the interaction features include the subspace blocks of overlapping actions and the docking envelope of the two agents in the interaction; If there is no interaction between any two agents, the interaction feature is an empty set.

4. A multi-agent collaborative interaction method according to claim 2, characterized in that: The actions are split layer by layer based on the goal or sub-goal until one or more time-series-based action sets to be executed are obtained; and the agent set is obtained based on the action set to be executed and the spatial range involved in the goal.

5. A multi-agent collaborative interaction method according to claim 4, characterized in that: Based on the target or sub-target, traverse the agents according to the distance, and take the agent that is closest to the target or sub-target and matches the last sequential action to be executed in each set of actions to be executed, and use this agent as the trigger agent corresponding to the current set of actions to be executed.

6. A multi-agent collaborative interaction method according to claim 5, characterized in that: After the agent is triggered to give the initial action, it is combined with its interactive features. Reversely obtain one or more actions that drive the front-end agent, repeat until the derivation of each set of actions to be executed is completed, calculate the cost based on the docking envelope position of each two agents, and obtain the optimal decision.

7. A multi-agent collaborative interaction method according to claim 6, characterized in that: Based on the optimal decision, the interactive information between the docking envelope positions of the two intelligent agents and the three-dimensional environment is obtained, and multi-agent collaboration is performed based on the interactive information.

8. A computer-readable storage medium, characterized in that: A multi-agent collaborative interaction program is stored thereon, and when the program is executed by a processor, the multi-agent collaborative interaction method described in one of claims 1 to 7 is implemented.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, it implements the multi-agent collaborative interaction method described in one of claims 1 to 7.

10. A multi-agent collaborative interaction system, characterized in that: The system comprises: At least two agents, the agents are configured in a VR device environment; A controller is provided in conjunction with the two intelligent agents, and the controller executes the multi-agent collaborative interaction method described in one of claims 1 to 7.

Images