Multi-agent scheduling method, device and equipment based on path optimization, medium and program

Abstract

The embodiment of the invention discloses a multi-agent scheduling method, device, equipment, medium and program based on path optimization, and the method comprises the steps: obtaining team cooperation interaction data of a target team, and generating an agent interaction diagram according to the team cooperation interaction data; wherein the target team comprises a plurality of team members, and each team member represents one agent; optimizing and updating the agent interaction diagram according to target reference data in the team cooperation interaction data; accumulating and updating pheromones of each interactive cooperation path in the agent interaction graph by adopting an ant colony algorithm; selecting a target interactive cooperation path according to the pheromone concentration of each interactive cooperation path in the agent interaction diagram; and instructing the team members to cooperatively execute team tasks according to the target interactive cooperation path. According to the technical scheme of the embodiment of the invention, the intelligence and rationality of multi-agent scheduling can be improved.

Description

Technical Field

[0001] Embodiments of the present invention relate to the fields of intelligent scheduling and artificial intelligence technology, and in particular to a multi-agent scheduling method, device, electronic device, storage medium and program based on path optimization. Background Art

[0002] With the development of artificial intelligence and multi-agent systems, more and more application scenarios require multiple agents to work together to complete complex team tasks. Agents need to interact and collaborate efficiently to achieve the overall goal.

[0003] However, how to determine the optimal collaborative path to minimize task completion time or other costs is a challenging problem faced in multi-agent intelligent scheduling.

[0004] In the process of implementing the present invention, the inventors found that the prior art has the following defects: the traditional single-agent scheduling method is difficult to effectively handle the complex interaction relationship between multiple agents. Because in practical applications, the interaction time between agents may be affected by various factors, such as network delay and resource competition. Therefore, a method that can consider the interaction relationship between agents and find the global optimal collaboration path is needed to realize the intelligent scheduling of multiple agents. Summary of the invention

[0005] The embodiments of the present invention provide a multi-agent scheduling method, device, electronic device, storage medium and program based on path optimization, which can improve the intelligence and rationality of multi-agent scheduling.

[0006] According to one aspect of the present invention, a multi-agent scheduling method based on path optimization is provided, comprising:

[0007] Acquire team collaboration interaction data of a target team, and generate an agent interaction graph according to the team collaboration interaction data; wherein the target team includes a plurality of team members, and each team member represents an agent;

[0008] Optimizing and updating the agent interaction graph according to the target reference data in the team collaboration interaction data;

[0009] An ant colony algorithm is used to accumulate and update the pheromones of each interactive cooperation path in the agent interaction graph;

[0010] Selecting a target interactive collaboration path according to the pheromone concentration of each interactive collaboration path in the agent interaction graph;

[0011] Instruct each of the team members to collaboratively perform a team task according to the target interactive collaboration path.

[0012] According to another aspect of the present invention, a multi-agent scheduling device based on path optimization is provided, comprising:

[0013] An agent interaction graph generation module is used to obtain team collaboration interaction data of a target team and generate an agent interaction graph according to the team collaboration interaction data; wherein the target team includes a plurality of team members, each of which represents an agent;

[0014] An interaction graph optimization and updating module, used to optimize and update the agent interaction graph according to the target reference data in the team collaboration interaction data;

[0015] A pheromone updating module, used for accumulating and updating the pheromones of each interactive cooperation path in the agent interaction graph by using an ant colony algorithm;

[0016] A target interactive collaboration path selection module, used for selecting a target interactive collaboration path according to the pheromone concentration of each interactive collaboration path in the agent interaction graph;

[0017] The team task execution module instructs each of the team members to collaboratively execute the team task according to the target interactive collaboration path.

[0018] According to another aspect of the present invention, an electronic device is provided, the electronic device comprising:

[0019] at least one processor; and

[0020] a memory communicatively connected to the at least one processor; wherein,

[0021] The memory stores a computer program that can be executed by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute the multi-agent scheduling method based on path optimization described in any embodiment of the present invention.

[0022] According to another aspect of the present invention, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a processor to implement the multi-agent scheduling method based on path optimization described in any embodiment of the present invention when executed.

[0023] According to another aspect of the present invention, a computer program product is provided, including a computer program, which, when executed by a processor, implements the multi-agent scheduling method based on path optimization described in any embodiment of the present invention.

[0024] The embodiment of the present invention obtains the team collaboration interaction data of a target team including multiple team members, generates an agent interaction graph based on the team collaboration interaction data, and regards each team member of the target team as an agent. Furthermore, the agent interaction graph is optimized and updated according to the target reference data in the team collaboration interaction data, and then the ant colony algorithm is used to accumulate and update the pheromones of each interactive collaboration path in the agent interaction graph, and finally the target interactive collaboration path is selected according to the pheromone concentration of each interactive collaboration path in the agent interaction graph. After determining the target interactive collaboration path, each team member can be instructed to collaborate to perform team tasks according to the target interactive collaboration path. The above technical solution can solve the problem of unreasonable scheduling methods in existing multi-agent scheduling methods, and can improve the intelligence and rationality of multi-agent scheduling.

[0025] It should be understood that the contents described in this section are not intended to identify the key or important features of the embodiments of the present invention, nor are they intended to limit the scope of the present invention. Other features of the present invention will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0027] Figure 1 is a flow chart of a multi-agent scheduling method based on path optimization provided by Embodiment 1 of the present invention;

[0028] Figure 2 is a flow chart of a multi-agent scheduling method based on path optimization provided by Embodiment 2 of the present invention;

[0029] Figure 3 It is a flowchart of a multi-agent scheduling system executing a multi-agent scheduling method provided by Embodiment 2 of the present invention;

[0030] Figure 4 is a schematic diagram of the structure of an intelligent agent interaction graph provided by Embodiment 2 of the present invention;

[0031] Figure 5 is a schematic diagram of a multi-agent scheduling device based on path optimization provided by Embodiment 3 of the present invention;

[0032] Figure 6 A schematic diagram of the structure of an electronic device provided in Embodiment 4 of the present invention. DETAILED DESCRIPTION

[0033] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0034] It should be noted that the terms "include", "have" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or apparatus comprising a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to these processes, methods, products or apparatus.

[0035] Embodiment 1

[0036] Figure 1 This is a flowchart of a multi-agent scheduling method based on path optimization provided in the first embodiment of the present invention. This embodiment can be applied to the situation where the target reference data in the team collaboration interaction data is used to optimize and update the agent interaction graph, and then the target interaction collaboration path of the multi-agents is determined by combining the updated agent interaction graph with the ant colony algorithm, and then the multi-agent scheduling is realized based on the target interaction collaboration path. This method can be executed by a multi-agent scheduling device based on path optimization. The device can be implemented by software and / or hardware, and can generally be integrated in an electronic device. The electronic device can be a terminal device or a server device. As long as it can execute the multi-agent scheduling method based on path optimization, the embodiment of the present invention does not limit the specific device type of the electronic device. Accordingly, if Figure 1 As shown, the method includes the following operations:

[0037] S110, obtaining team collaboration interaction data of a target team, and generating an agent interaction graph according to the team collaboration interaction data; wherein the target team includes a plurality of team members, and each team member represents an agent.

[0038] The target team may include multiple team members, and the team members of the target team may complete the team task together in a collaborative manner. The target team may be any type of team, and the team members in the target team may also be any type of members. The embodiment of the present invention does not limit the types of the target team and the team members constituting the target team. The team collaboration interaction data may be relevant data for the interaction between the team members in the target team. The agent interaction graph may be a schematic diagram of the interaction between the agents generated for the target team.

[0039] For example, the target team can be a research and development team, and the team members can be relevant research and development personnel in the research and development team, such as front-end engineers, back-end engineers, and test engineers. The target team can also be a logistics team, and the team members can be relevant logistics personnel in the logistics team, such as express collection personnel, express transport personnel, and express delivery personnel. The target team can also be a team composed of multiple departments, and the team members can be various departments, such as demand management department, research and development department, sales department, and after-sales service department. The target team can also be a team composed of multiple devices, and the team members can be various devices. For example, if the target team is a cloud computing cluster, the team members can be various node devices in the cloud computing cluster. A team member in the target team can represent an intelligent agent, and the interaction between team members can be reflected through the intelligent agent interaction diagram.

[0040] After determining the target team type and the team members in the target team, the collaborative interaction data between the team members can be collected as team collaborative interaction data, and the agent interaction graph of the target team can be constructed based on the acquired team collaborative interaction data. The agent interaction graph can reflect the real interaction situation between the team members.

[0041] Specifically, the agent interaction graph may include a path consisting of multiple nodes, each node represents an agent, i.e., a team member, and the nodes may include connecting edges (hereinafter referred to as edges). If there is an edge between two nodes, it means that the two nodes can interact in the direction guided by the edge. Multiple nodes and corresponding edges can form a complete path, and each path in the agent interaction graph can represent a feasible collaborative method for each team member to perform team tasks.

[0042] It is understandable that a target team can collaborate to complete one or more different types of team tasks, and the content of the team collaboration interaction data corresponding to different types of team tasks may also be different. Therefore, when generating an agent interaction graph based on the team collaboration interaction data, the team collaboration interaction data when the target team performs the target team task can be specifically obtained, and the agent interaction graph corresponding to the target team task can be generated based on the team collaboration interaction data when the target team performs the target team task. That is, one type of team task can generate one type of agent interaction graph. Different team tasks of a target team may generate different agent interaction graphs.

[0043] S120. Optimize and update the agent interaction graph according to the target reference data in the team collaboration interaction data.

[0044] Among them, the target reference data can be reference data in the team collaboration interaction data that can optimize and update the constructed intelligent agent interaction graph.

[0045] It is understandable that the agent interaction diagram generated based on the team collaboration interaction data can reflect the actual interaction between the team members in the initial state, but may not reflect the quality of interaction between the team members. For example, although both paths A→B→C and A→D→C can complete the team task of the target team, the time cost or economic cost required to complete the team task by the two paths is different. Because different types of team tasks may focus on different task completion indicators, for example, team task A requires the time cost to be as low as possible, team task B requires the economic cost to be as low as possible, and team task C requires the task completion quality to be as high as possible. Therefore, after obtaining the agent interaction diagram, it is also necessary to screen the target reference data from the team collaboration interaction data according to the type of task completion indicator focused on by the team task, and optimize and update the agent interaction diagram according to the target reference data.

[0046] Exemplarily, the interaction results between agents (such as successful interaction or failed interaction, etc.) can be used as target reference data, and the edges in the agent interaction graph can be adjusted according to the interaction results between the agents. For example, the weights of the edges can be adjusted, or the edges in the agent interaction graph can be deleted or added.

[0047] Exemplarily, the interaction costs between agents (such as economic costs, etc.) can also be used as target reference data, and the edges in the agent interaction graph can be adjusted according to the interaction costs between the agents. For example, the weights of the edges can be adjusted, or the edges in the agent interaction graph can be deleted or added.

[0048] It is understandable that different types of target teams and / or team tasks can be configured with different target reference data, and the embodiment of the present invention does not limit the specific type of target reference data. Accordingly, the final agent interaction diagram obtained by optimizing and updating the agent interaction diagram according to the target reference data can meet the team task indicator requirements or meet the actual task execution situation, that is, the composition structure of the agent interaction diagram is more scientific and reasonable.

[0049] S130, using an ant colony algorithm to accumulate and update the pheromones of each interactive collaboration path in the agent interaction graph.

[0050] Among them, the interactive collaboration path is a path composed of nodes and corresponding edges in the agent interaction graph. An interactive collaboration path represents a feasible collaborative way for multiple team members to perform team tasks. Pheromone is the core concept of the ant colony algorithm. It is a virtual chemical signal used to simulate the secretions left by real ants in the process of searching for food. In the ant colony algorithm, pheromone is used to indicate the quality of the path. That is, the higher the concentration of pheromone on the path, the more likely it is that the path is a high-quality path.

[0051] The ant colony algorithm has strong robustness and good distributed computing characteristics. It can simulate the foraging behavior of ants and find the optimal path through the accumulation and update of pheromones, which is consistent with the physical process of path planning. This makes the ant colony algorithm very suitable for application in path planning scenarios. Therefore, in an embodiment of the present invention, when an agent interaction graph is generated according to the team collaboration interaction data, and the target reference data in the team collaboration interaction data is further used to optimize and update the agent interaction graph, and an agent interaction graph that is more in line with the actual situation is obtained, it can be further combined with the ant colony algorithm, with the agent interaction graph as a benchmark, and according to the calculation logic of the ant colony algorithm, the pheromones of each interactive collaboration path in the agent interaction graph are accumulated and updated.

[0052] S140: Select a target interactive collaboration path according to the pheromone concentration of each interactive collaboration path in the agent interaction graph.

[0053] The target interactive collaboration path may be the optimal interactive collaboration path selected from various interactive collaboration paths of the agent interaction graph.

[0054] It is understandable that different interactive collaboration paths can calculate different pheromone concentration values. Since the higher the pheromone concentration, the higher the quality of the interactive collaboration path. Therefore, after the final pheromone concentration is calculated for each interactive collaboration path in the agent interaction graph, the interactive collaboration path with the highest pheromone concentration can be selected as the target interactive collaboration path.

[0055] S150: Instruct each of the team members to collaboratively perform a team task according to the target interactive collaboration path.

[0056] Correspondingly, after determining the target interactive collaboration path, each team member can be instructed to collaboratively execute the team task corresponding to the agent interaction diagram according to the target interactive collaboration path.

[0057] Exemplarily, assume that the target interactive collaboration path is A→D→C, where A, D and C are three different computer devices. When instructing each team member to collaborate in performing a team task according to the target interactive collaboration path, an instruction can be first issued to computer device A, and computer device A can first complete part of the team task. When computer device A completes the corresponding part of the team task, the relevant task data can be sent to computer device D. Correspondingly, computer device D can further complete the part of the team task that needs to be completed by this device based on the task data sent by computer device A. When computer device D completes the corresponding part of the team task, the relevant task data can be sent to computer device C. Correspondingly, computer device C can further complete the part of the team task that needs to be completed by this device based on the task data sent by computer device D. When computer device C completes the corresponding part of the team task, the entire team task is completed.

[0058] The embodiment of the present invention obtains the team collaboration interaction data of a target team including multiple team members, generates an agent interaction graph based on the team collaboration interaction data, and regards each team member of the target team as an agent. Furthermore, the agent interaction graph is optimized and updated according to the target reference data in the team collaboration interaction data, and then the ant colony algorithm is used to accumulate and update the pheromones of each interactive collaboration path in the agent interaction graph, and finally the target interactive collaboration path is selected according to the pheromone concentration of each interactive collaboration path in the agent interaction graph. After determining the target interactive collaboration path, each team member can be instructed to collaborate to perform team tasks according to the target interactive collaboration path. The above technical solution can solve the problem of unreasonable scheduling methods in existing multi-agent scheduling methods, and can improve the intelligence and rationality of multi-agent scheduling.

[0059] Embodiment 2

[0060] Figure 2 is a flow chart of a multi-agent scheduling method based on path optimization provided by Embodiment 2 of the present invention. Figure 3 is a flowchart of a multi-agent scheduling system executing a multi-agent scheduling method provided by the second embodiment of the present invention. This embodiment is specific based on the above embodiment. In this embodiment, multiple specific optional implementation methods are provided for generating an agent interaction graph based on team collaboration interaction data, optimizing and updating the agent interaction graph, and accumulating and updating the pheromones of each interactive collaboration path in the agent interaction graph. Figure 2 As shown, the method of this embodiment may include:

[0061] S210. Acquire team collaboration interaction data of a target team, wherein the target team includes a plurality of team members, and each team member represents an intelligent agent.

[0062] Optionally, the team collaboration interaction data may include agent identification, interaction agent object identification, interaction start cost, interaction end cost and interaction result.

[0063] Among them, the agent identifier can be used to uniquely identify the agent. Exemplarily, the agent identifier can be an agent ID (Identification number). The interactive agent object identifier can be a unique identifier of another agent that interacts with the current agent. Exemplarily, the interactive agent object identifier can be an interactive agent object ID. The interaction start cost can be the cost before the two agents interact, and the interaction end cost can be the cost spent after the two agents complete the interaction. Optionally, the interaction cost can include but is not limited to the interaction time, interaction labor cost and interaction economic cost, etc., and the embodiment of the present invention does not limit the specific type of interaction cost. Exemplarily, the interaction start cost can be the interaction start time, and the interaction end cost can be the interaction end time. Among them, the interaction start time can be the time point when the two agents start to interact. The interaction end time can be the time point when the two agents end the interaction. The interaction result can be the result of the interaction between the two agents, for example, it can include but is not limited to the result types such as successful interaction or failed interaction. Optionally, the interaction result of successful interaction or failed interaction can be further refined, for example, it can include successful interaction with low interaction quality, successful interaction with high interaction quality, and failed interaction but short interaction time. That is, the type of interaction result can be configured according to specific task requirements, and the embodiment of the present invention does not limit the result type corresponding to the interaction result.

[0064] In a specific example, Figure 3As shown, the entire multi-agent scheduling method can be executed by the multi-agent scheduling system. Optionally, the multi-agent scheduling system may include an intelligent scheduling module, which can receive the collaborative tasks submitted by the user, so as to determine the team tasks that each team member in the target team needs to perform collaboratively. Accordingly, after the intelligent scheduling module determines the team tasks, it can assign detailed tasks to each agent. The agent can perform the tasks assigned by the intelligent scheduling module and interact with other agents in the process of performing the tasks. During the interaction process of each agent, each agent can record the interaction log and send the interaction log data to the log database in the multi-agent scheduling system for storage. Further, the intelligent scheduling module can read the interaction log from the log database, generate an agent interaction graph based on the interaction log, and call the ant colony algorithm module in the multi-agent scheduling system, and send the agent interaction graph generated according to the log interaction data to the ant colony algorithm module. The ant colony algorithm module executes the path optimization algorithm according to the agent interaction graph, and the optimal path in the screening is fed back to the intelligent scheduling module. Further, the intelligent scheduling module can feed back the screened optimal path to the user for display. At the same time, the multi-agent scheduling system can schedule each related agent to perform tasks in sequence according to the optimal collaborative path and feedback the results.

[0065] Exemplarily, the teamwork interaction data may be interaction log data in a specific data format recorded by an agent, which may include but is not limited to agent identification, interactive agent object identification, interaction start time, interaction end time, and interaction results. Among them, the agent identification may record the unique identifier of the agent, such as named agent_id. The interactive agent object identification may record the unique identifier of another agent interacting with the current agent, such as named interaction_agent_id. The interaction start time may record the precise timestamp of the start of the interaction, such as named start_time. The interaction end time may record the precise timestamp of the end of the interaction, such as named end_time. The interaction result may record the result of the agent interaction, such as named interaction_result. The interaction result type may include success or failure, as well as specific completed tasks, which is conducive to more complex scenario analysis and algorithm improvement.

[0066] In a specific example, team collaboration interaction data may include the following:

[0067] agent_id,interaction_agent_id,start_time,end_time,interaction_result

[0068] agent_1,agent_2,2024-10-27 10:00:00,2024-10-27 10:05:00,success

[0069] agent_2,agent_3,2024-10-27 10:05:00,2024-10-27 10:10:00,success

[0070] agent_1,agent_3,2024-10-27 10:10:00,2024-10-27 10:12:00,failed

[0071] In order to ensure the accuracy and reliability of the team collaboration interaction data, some rules can be configured for the team collaboration interaction data. Exemplarily, the rules configured for the team collaboration interaction data may include the following:

[0072] (1) All timestamps should use a unified standard format, such as the ISO 8601 format, i.e. YYYY-MM-DD HH:mm:ss.

[0073] (2) Agent identifiers and interactive agent object identifiers should be unique identifiers.

[0074] (3) Each interaction between agents should record a line of data.

[0075] (4) If the interaction between agents fails, the actual end time of the interaction should still be recorded.

[0076] (5) The fields of interaction results are optional and can be adjusted according to actual needs.

[0077] S220. Generate a node in the agent interaction graph according to the agent identifier and the interactive agent object identifier.

[0078] Accordingly, after obtaining the teamwork interaction data, it can be preprocessed for algorithm calculation. After the teamwork interaction data is preprocessed, the nodes of the agent interaction graph can be constructed according to the agent identifier and the interaction agent object identifier. Optionally, the agent interaction graph can be represented as G=(V,E). Among them, the node V represents the agent, and the edge E represents the interaction between the agents.

[0079] S230. Calculate the node interaction cost according to the interaction start cost and the interaction end cost, and generate the weight of the edge between each node in the agent interaction graph according to the node interaction cost.

[0080] The node interaction cost may be the total cost required for two nodes to complete the interaction.

[0081] Optionally, the difference between the interaction end cost and the interaction start cost corresponding to the interaction action between two nodes can be calculated to obtain the node interaction cost between the two nodes, and the node interaction cost can be used as the weight w of the edge between each node in the agent interaction graph ij .

[0082] For example, taking the interaction time as the interaction cost, the interaction duration can be calculated according to the interaction start time and the interaction end time, that is, the difference between the interaction end time and the interaction start time corresponding to the interaction action between the two nodes is calculated to obtain the interaction duration between the two nodes, and the weight w of the edge between each node in the agent interaction graph is generated according to the interaction duration. ij The interaction duration may be the duration of the interaction action between two agents.

[0083] Exemplarily, taking the interaction economic cost as an example of the interaction cost, the total interaction economic cost between two nodes can be calculated based on the interaction start economic cost and the interaction end economic cost, that is, the difference between the interaction end economic cost and the interaction start economic cost corresponding to the interaction action between the two nodes is calculated to obtain the economic cost required for the interaction between the two nodes, and the weight w of the edge between each node in the agent interaction graph is generated based on the economic cost required for the interaction between the two nodes. ij .

[0084] In a specific example, suppose there are two agents, agent_1 and agent_2, and they interact with each other. The team collaboration interaction data generated is log data, and the data recording format is: agent_id, interaction_agent_id, start_time, end_time, interaction_result. The actual log data is: agent_1, agent_2, 2024-10-27 10:00:00, 2024-10-27 10:02:00, success. Furthermore, the following information can be extracted from the log data: the two parties in the interaction: agent_1 and agent_2; the duration of the interaction: 2 minutes. Based on the log data, an edge between the nodes agent_1 and agent_2 in the agent interaction graph can be calculated: agent_1--(2)-->agent_2, where the value 2 on the edge represents the weight of the edge, that is, the duration of the interaction. In this way, by collecting and processing these team collaboration interaction data, we can construct an agent interaction graph, and use algorithms to find the optimal collaboration path based on the agent interaction graph, thereby optimizing the collaboration efficiency of multiple agents.

[0085] S240: Determine the type of interaction result between the nodes according to the team collaboration interaction data.

[0086] S250: Optimize and update the weights of the edges according to the types of interaction results between the nodes.

[0087] It is understandable that the simple node interaction cost may not fully and accurately reflect the cost of the interaction action. For example, a failed interaction may require correction or compensation of subsequent interaction actions, and the overall actual cost is higher than a successful interaction, even if the two take the same time. Therefore, the interaction results can be used as target reference data, and the types of interaction results between each agent can be determined, and then the weights of the edges can be optimized and updated according to the types of interaction results between each node.

[0088] In an optional embodiment of the present invention, the optimizing and updating of the weights of the edges according to the types of interaction results between the nodes may include: when it is determined that the type of interaction result is a failed interaction result, increasing the weight value of the edge corresponding to the failed interaction result; when it is determined that the type of interaction result is a successful interaction result, reducing the weight value of the edge corresponding to the successful interaction result.

[0089] It can be understood that the larger the weight value of the edge between nodes, the larger the interaction cost between the two nodes. The embodiment of the present invention uses the node interaction cost as the weight value of the edge between nodes. Therefore, in the embodiment of the present invention, the larger the weight value of the edge between nodes, the longer the node interaction cost required between the two nodes. Optionally, if the interaction result type only includes two types of successful interaction and failed interaction, when optimizing and updating the weight of the edge according to the interaction result type between each node, for the type of failed interaction result, the weight value of the corresponding edge can be increased to increase the node interaction cost between the corresponding nodes so that it can be avoided as much as possible in the path selection; for the type of successful interaction result, the weight value of the corresponding edge can be reduced to reduce the node interaction cost between the corresponding nodes so that it can be selected as much as possible in the path selection.

[0090] Optionally, the types of failed interaction results and successful interaction results can be further divided into finer and more hierarchical types, and a method for updating edge weight values ​​can be set for each interaction result type. For example, the weight of the edge corresponding to the type of high-quality successful interaction results can be assigned a lower weight value; the weight of the edge corresponding to the type of low-quality successful interaction results can be assigned a slightly higher weight value, but lower than the weight value of the edge corresponding to the type of failed interaction results. Updating the edge weight value according to the multi-level interaction result type can further improve the rationality and accuracy of the edge weight value configuration, which is not limited in the embodiments of the present invention.

[0091] S260. Optimize and update the edges in the agent interaction graph according to the types of interaction results between the nodes.

[0092] In some scenarios, the results of agent interactions may affect the possibility of subsequent interactions. For example, if an interaction fails, some subsequent interactions may not be possible. Therefore, in order to ensure the rationality of the agent interaction graph structure, the structure of the edges in the agent interaction graph can also be optimized and updated according to the types of interaction results between nodes, thereby achieving the optimization update of the agent interaction graph structure.

[0093] In an optional embodiment of the present invention, the optimizing and updating of the edges in the agent interaction graph according to the interaction result type between each of the nodes may include: when it is determined that the interaction result type is a failed interaction result, and when it is determined that the edge corresponding to the failed interaction result is an infeasible edge, removing the infeasible edge from the agent interaction graph; and when it is determined that the interaction result type is converted from a failed interaction result to a successful interaction result, adding edges between nodes converted from failed interaction results to successful interaction results.

[0094] Optionally, if the interaction result types only include two types: successful interaction and failed interaction, when optimizing and updating the edges in the agent interaction graph according to the interaction result types between the nodes, for the failed interaction result, it is possible to further determine whether the edge corresponding to the failed interaction result is an infeasible edge. If it is determined that the edge corresponding to the failed interaction result is an infeasible edge, the edge that has become infeasible due to the previous interaction failure can be removed. In the case where the failed interaction result is converted to a successful interaction result again, a new edge can be added between the nodes where the failed interaction result is converted to a successful interaction result again.

[0095] The above technical solution can further improve the performance and adaptability of the ant colony algorithm by improving the edge weights of the agent interaction graph by utilizing the interaction results and dynamically adjusting the graph structure of the agent interaction graph, so that it can better cope with complex situations in practical applications.

[0096] S270, using an ant colony algorithm to accumulate and update the pheromones of each interactive collaboration path in the agent interaction graph.

[0097] In an optional embodiment of the present invention, the use of an ant colony algorithm to accumulate and update the pheromones of each interactive collaboration path in the agent interaction graph may include: initializing the algorithm parameters of the ant colony algorithm; wherein the algorithm parameters of the ant colony algorithm include the number of ants, pheromone volatility coefficient, pheromone importance factor, heuristic information importance factor, maximum number of iterations, and pheromone concentration of the edges in the agent interaction graph; determining the starting node and the ending node in the agent interaction graph; placing each ant at the starting node, and each ant selects the next node in the agent interaction graph according to the state transition probability until reaching the ending node; and updating the pheromone concentration of the edges in the agent interaction graph according to the pheromone increment accumulated by each ant on the edge.

[0098] In a specific example, the collaborative process of the agents can be modeled as a directed weighted graph G=(V,E) as the agent interaction graph, where V represents the node set, representing the agents participating in the collaboration, and E represents the edge set, representing the interaction relationship between the agents. The weight w of the edge (i,j) is ij Represents the time required for agent i to interact with agent j. The goal of the ant colony algorithm is to find a path from the starting node s to the target node t based on the agent interaction graph, so that the target indicator on the path is minimized. Exemplarily, the target indicator can be, for example, total time, total labor cost or total economic cost, etc., which can be configured according to the specific execution requirements of the team task. The embodiment of the present invention does not limit the type of target indicator referenced by the ant colony algorithm.

[0099] Correspondingly, when the agent interaction graph is constructed and the target reference data is used to complete the optimization and update process of the graph, the algorithm parameters of the ant colony algorithm can be initialized, that is, the initialization values ​​of the algorithm parameters such as the number of ants, pheromone volatility coefficient, pheromone importance factor, heuristic information importance factor, maximum number of iterations, and pheromone concentration of edges in the agent interaction graph can be set. Exemplarily, the number of ants can be set to m, the pheromone volatility coefficient can be set to ρ (0<ρ<1), the pheromone importance factor (also called pheromone importance factor) can be set to α, the heuristic information importance factor (also called heuristic function importance factor) can be set to β, and the maximum number of iterations can be set to MaxIter. At the same time, the pheromone concentration τ of all edges in the agent interaction graph can be initialized ij is a small positive value.

[0100] Furthermore, the starting node s and the ending node t are selected from the agent interaction graph. For each ant k, it can be placed at the starting node s. Each ant can select the next node in the agent interaction graph based on the current node according to the state transition probability until it reaches the ending node t.

[0101] In an optional embodiment of the present invention, the formula of the state transition probability refers to the following expression:

[0102]

[0103] Among them, P ij k represents the state transition probability of the ant from the current node i to the next node j, τ ij represents the pheromone concentration of the edge between node i and node j, η ij represents the expected degree from node i to node j, allowed_k represents the next node set selected by ant k, τ il represents the pheromone concentration of the edge between node l and node j, η il represents the expected degree from node l to node j, α represents the pheromone importance factor, β represents the heuristic information importance factor, and l represents a natural number.

[0104] Each ant will increase the pheromone increment on each path from the starting node s to the ending node t, and then update the pheromone concentration of the edge in the agent interaction graph according to the pheromone increment. It can be understood that the pheromone increment is the increase in the amount of pheromone.

[0105] In an optional embodiment of the present invention, updating the pheromone concentration of the edge in the agent interaction graph according to the pheromone increment accumulated by each ant on the edge may include: updating the pheromone concentration of the edge in the agent interaction graph according to the pheromone increment based on the following formula:

[0106]

[0107] Wherein, ρ represents the volatility coefficient of the pheromone, Represents the pheromone increment of ant k on the edge between node i and node j.

[0108] If ant k passes through the edge between node i and node j, then =Q / L k , if ant k does not pass through the edge between node i and node j, then = 0, Q is the pheromone intensity constant, L k is the length of the path traversed by ant k. For example, if the shortest total time is the path optimization goal, then L k It can be the total time of the path traversed by ant k.

[0109] Each ant repeats the above operation, that is, repeatedly placing each ant on the starting node, each ant selects the next node in the agent interaction graph according to the state transition probability until it reaches the end node, and updates the pheromone concentration of the edge in the agent interaction graph according to the pheromone increment accumulated by each ant on the edge until it is determined that the maximum number of iterations has been reached.

[0110] S280: Select a target interactive collaboration path according to the pheromone concentration of each interactive collaboration path in the agent interaction graph.

[0111] S290: Instruct each of the team members to collaboratively perform a team task according to the target interactive collaboration path.

[0112] It should be noted that Figure 2 This is only a schematic diagram of an implementation method, and there is no order of execution between step S250 and step S260. That is, step S250 may be executed first, and then step S260, or step S260 may be executed first, and then step S250, or both may be executed in parallel or one may be executed selectively, and the embodiment of the present invention does not limit this.

[0113] In an optional embodiment of the present invention, after selecting the target interactive collaboration path according to the pheromone concentrations of each interactive collaboration path in the agent interaction graph, it may also include: calculating the interaction failure rate of the target interactive collaboration path; when it is determined that the interaction failure rate of the target interactive collaboration path exceeds a preset failure rate threshold, adjusting the algorithm parameters of the ant colony algorithm; and updating the pheromone calculation method in the ant colony algorithm according to the target reference data.

[0114] The interaction failure rate may be a ratio between the number of interaction failure results and the total number of interactions. The preset failure rate threshold may be set according to requirements, and the embodiment of the present invention does not limit the specific value of the preset failure rate threshold.

[0115] In the embodiment of the present invention, the interaction result can also be used as an important indicator for evaluating the performance of the ant colony algorithm. By analyzing the distribution of the interaction results on different paths, the effectiveness of the ant colony algorithm can be evaluated, and the improvement and parameter optimization of the ant colony algorithm can be guided.

[0116] In a specific example, if it is found that the interaction failure rate on the shortest path (i.e., the target interactive collaboration path) screened by the ant colony algorithm is very high, or even exceeds the preset failure rate threshold, it may be necessary to adjust the algorithm parameters of the ant colony algorithm, such as adjusting the pheromone volatility coefficient or the heuristic information importance factor, to pay more attention to the success rate of the interaction.

[0117] Optionally, a more complex reward mechanism suitable for the ant colony algorithm can be designed based on target reference data such as interaction results, etc., to update the pheromone calculation method in the ant colony algorithm, thereby guiding the ant colony algorithm to screen out more reliable paths.

[0118] Figure 4 is a schematic diagram of the structure of an agent interaction diagram provided by the second embodiment of the present invention. In a specific example, Figure 4 As shown in the figure, it is assumed that there are four agents A, B, C and D. The numbers in the brackets represent the node interaction costs such as the duration of the interaction, that is, the weight of the edge. The goal of the team task is to find the shortest path from A to D, so as to find the path with the shortest total time required from A to D. Initialize the algorithm parameters of the ant colony algorithm, set the number of ants m=2, the pheromone volatility coefficient ρ=0.5, the pheromone importance factor α=1, the heuristic information importance factor β=1, the pheromone intensity constant Q=1, and the maximum number of iterations MaxIter=2. After two iterations, the pheromone concentration of the path A->B->D is the highest: τ AB +τ BD =1.5, so it is selected as the optimal path as the target interactive collaboration path.

[0119] It should be noted that the above example is simplified to illustrate the calculation process. In actual applications, there will be more ants, more iterations, and more accurate calculations of state transition probabilities, resulting in more accurate results. In optimizing the above algorithm, the interaction results can be used to adjust the weights of the edges in the agent interaction graph. For example, if the interaction from B->D often fails, the weight of BD can be increased from 1 to 10, etc., so that the ant colony algorithm tends to choose other paths.

[0120] The embodiment of the present invention provides a multi-agent collaborative path optimization method based on an ant colony algorithm. The ant colony algorithm uses the accumulation and update mechanism of pheromones in the ant colony algorithm and based on the log report data of the agent, the optimal path is found in the agent interaction graph, thereby minimizing the reference index of task completion, such as minimizing the time cost or economic cost of task completion. The ant colony algorithm has a strong global search capability and can effectively explore the solution space to avoid falling into the local optimal solution, which is crucial for complex multi-agent collaborative scenarios. At the same time, the ant colony algorithm can adapt to dynamic environmental changes. Through the pheromone update mechanism, the path selection can be adjusted according to the real-time interaction cost, thereby coping with the influence of dynamic factors such as network delay and resource competition. The ant colony algorithm is less sensitive to parameter changes and has a certain robustness. Even if the parameter setting is not optimal, the algorithm can still find a better solution. At the same time, the ant colony algorithm can be easily parallelized, thereby improving the computing efficiency, and can improve the efficiency of agent scheduling in large-scale multi-agent collaborative scenarios. Therefore, by comprehensively utilizing the agent interaction graph that is optimized and updated according to the target reference data in the teamwork interaction data, and combining it with the ant colony algorithm to realize multi-agent scheduling, it is possible to handle the complex interaction relationship between multiple agents, adapt well to the dynamic interaction environment, and not easily fall into the local optimal solution. The multi-agent scheduling method based on path optimization provided by the embodiment of the present invention can be easily extended to more complex scenarios, such as considering factors such as agent capability differences and task priorities, and by introducing auxiliary information such as interaction results, the performance and adaptability of the algorithm can be further improved, so that it can better cope with complex situations in practical applications.

[0121] Specific application scenarios

[0122] In order to more clearly illustrate the technical solution provided by the embodiment of the present invention, the embodiment of the present invention takes the software development team as the target team as an example to illustrate how to determine the collaboration mode between team members in the target team based on the multi-agent scheduling method based on path optimization in the software development scenario, so as to collaboratively execute the team task of completing software development.

[0123] Assume that a software development team has three members: front-end engineer (abbreviated as FE), back-end engineer (abbreviated as BE), and test engineer (abbreviated as TE). They need to collaborate to complete the development of a new software function. Obtain the team collaboration interaction data of each team member, including the agent identifier (which can represent the initiator), the interaction agent object identifier (which can represent the recipient), the interaction start time, the interaction end time, and the interaction result. Calculate the interaction duration (referred to as time in the following table) between the initiator and the recipient based on the interaction start time and the interaction end time. At the same time, the interaction result types of success / failure / partial success are represented by S (success) / F (failure) / P (partial success), and the structured team collaboration interaction data shown in Table 1 is obtained:

[0124] Table 1 Structured team collaboration interaction data list

[0125]

[0126] Initialize the ant colony algorithm parameters: the number of ants is set to 50; the pheromone volatility coefficient is set to 0.5; the pheromone importance factor is set to 1, the heuristic information importance factor is set to 2 (more emphasis on time cost), the pheromone intensity constant is set to 1, and the maximum number of iterations is set to 100.

[0127] Perform data preprocessing, that is, convert the interaction duration and interaction result data into a format that can be used by the ant colony algorithm, and construct a weighted agent interaction graph. The initial weight value of the edge of the agent interaction graph is the interaction duration.

[0128] Further, the pheromone is initialized, and the pheromone of all edges in the agent interaction graph is initialized to 1. After the pheromone initialization is completed, the iterative search process can be started. During the iterative search process, each ant selects a path based on the current pheromone concentration and distance (time). The length of the path found by each ant is calculated, and the path weight is adjusted according to the interaction result. Specifically, when the interaction result is failure (F), the weight of the edge containing the failed interaction is greatly increased, for example, multiplied by a penalty coefficient, such as setting the penalty coefficient to 3, etc., to simulate the rework cost. For example, when the interaction result between BE->FE is failure, the weight of the edge of BE->FE is increased from 2 to 2*3=6. When the interaction result is partial success (P), the weight of the edge containing the partially successful interaction is moderately increased, for example, by a fixed value of 1, etc., to simulate the cost of additional communication and adjustment. For example, when the interaction result between FE->TE is partially successful, the weight of the edge of FE->TE is increased from 2 to 2+1=3, and the weight of TE->BE is increased from 2 to 2+1=3.

[0129] When the weights of the agent interaction graph edges are optimized and updated, the pheromone concentration can be updated according to the adjusted path length in the agent interaction graph. In this scenario example, it is assumed that the communication channels between team members are fixed, so there is no need to dynamically adjust the graph structure of the agent interaction graph. If the impact of the interaction results is not considered, the ant colony algorithm may choose the path FE->BE->TE because its initial time cost is the lowest, 3+4=7. However, due to the failure of the BE->FE interaction, the actual required interaction time cost is much higher than 7. After adjusting the weights of the edges of the agent interaction graph according to the interaction results, the weight of BE->FE will increase significantly. Accordingly, the ant colony algorithm will tend to choose other paths, such as FE->TE->BE, even though the initial time cost of this path is higher, 2+4=6, but because the interaction result of TE->BE is partially successful, the time cost required for the TE->BE path after the weight adjustment is 6+1=7, which is lower than the adjusted cost of FE->BE->TE path, 3+6=9.

[0130] In summary, by introducing the interaction results to optimize and update the weights of the agent interaction graph edges, the collaboration cost can be calculated more precisely, not just limited to time, but also more reflective of hidden costs such as rework and communication in actual work. At the same time, by introducing the interaction results to optimize and update the weights of the agent interaction graph edges, the actual team collaboration scenario can be simulated more accurately, making the path found by the ant colony algorithm more practical. By guiding the ant colony algorithm to avoid high-risk interactions, such as interactions that often fail, the team's collaboration efficiency can be effectively improved. The above example scenario highlights how the interaction results affect the path selection of the ant colony algorithm, and ultimately helps find the optimal collaboration path that is more in line with the actual situation, showing how to use richer contextual information to improve the performance of the algorithm and make it more suitable for complex real-world scenarios.

[0131] It should be noted that the relevant information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data for display, data for analysis, etc.) involved in this disclosure are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data comply with relevant laws, regulations and standards in relevant regions.

[0132] It should be noted that any arrangement and combination of the technical features in the above embodiments also falls within the protection scope of the present invention.

[0133] Embodiment 3

[0134] Figure 5 is a schematic diagram of a multi-agent scheduling device based on path optimization provided by Embodiment 3 of the present invention. Figure 5As shown, the device includes: an agent interaction graph generation module 310, an interaction graph optimization and updating module 320, a pheromone updating module 330, a target interaction and collaboration path selection module 340 and a team task execution module 350, wherein:

[0135] An agent interaction graph generation module 310 is used to obtain team collaboration interaction data of a target team and generate an agent interaction graph according to the team collaboration interaction data; wherein the target team includes a plurality of team members, each of which represents an agent;

[0136] An interaction graph optimization and updating module 320, configured to optimize and update the agent interaction graph according to target reference data in the team collaboration interaction data;

[0137] The pheromone updating module 330 is used to accumulate and update the pheromone of each interactive cooperation path in the agent interaction graph by using an ant colony algorithm;

[0138] A target interactive collaboration path selection module 340 is used to select a target interactive collaboration path according to the pheromone concentration of each interactive collaboration path in the agent interaction graph;

[0139] The team task execution module 350 instructs each of the team members to collaboratively execute the team task according to the target interactive collaboration path.

[0140] The embodiment of the present invention obtains the team collaboration interaction data of a target team including multiple team members, generates an agent interaction graph based on the team collaboration interaction data, and regards each team member of the target team as an agent. Furthermore, the agent interaction graph is optimized and updated according to the target reference data in the team collaboration interaction data, and then the ant colony algorithm is used to accumulate and update the pheromones of each interactive collaboration path in the agent interaction graph, and finally the target interactive collaboration path is selected according to the pheromone concentration of each interactive collaboration path in the agent interaction graph. After determining the target interactive collaboration path, each team member can be instructed to collaborate to perform team tasks according to the target interactive collaboration path. The above technical solution can solve the problem of unreasonable scheduling methods in existing multi-agent scheduling methods, and can improve the intelligence and rationality of multi-agent scheduling.

[0141] Optionally, the team collaboration interaction data includes an agent identifier, an interacting agent object identifier, an interaction start cost, an interaction end cost and an interaction result; the agent interaction graph generation module 310 is also used to: generate nodes in the agent interaction graph according to the agent identifier and the interacting agent object identifier; calculate the node interaction cost according to the interaction start cost and the interaction end cost; and generate the weights of the edges between the nodes in the agent interaction graph according to the node interaction cost.

[0142] Optionally, the interaction graph optimization and update module 320 is also used to: determine the type of interaction results between each of the nodes based on the team collaboration interaction data; optimize and update the weights of the edges based on the type of interaction results between each of the nodes; and optimize and update the edges in the agent interaction graph based on the type of interaction results between each of the nodes.

[0143] Optionally, the interaction graph optimization and update module 320 is also used to: when it is determined that the interaction result type is a failed interaction result, increase the weight value of the edge corresponding to the failed interaction result; when it is determined that the interaction result type is a successful interaction result, reduce the weight value of the edge corresponding to the successful interaction result.

[0144] Optionally, the interaction graph optimization and update module 320 is also used to: when it is determined that the interaction result type is a failed interaction result, and when it is determined that the edge corresponding to the failed interaction result is an infeasible edge, remove the infeasible edge from the agent interaction graph; when it is determined that the interaction result type is converted from a failed interaction result to a successful interaction result, add edges between nodes that are converted from failed interaction results to successful interaction results.

[0145] Optionally, the pheromone update module 330 is also used to: initialize the algorithm parameters of the ant colony algorithm; wherein the algorithm parameters of the ant colony algorithm include the number of ants, pheromone volatility coefficient, pheromone importance factor, heuristic information importance factor, maximum number of iterations and pheromone concentration of the edges in the agent interaction graph; determine the starting node and the ending node in the agent interaction graph; place each ant at the starting node, and each ant selects the next node in the agent interaction graph according to the state transition probability until reaching the ending node; update the pheromone concentration of the edges in the agent interaction graph according to the pheromone increment accumulated by each ant on the edge.

[0146] Optionally, the above-mentioned device also includes an algorithm updating module, which is used to: calculate the interaction failure rate of the target interactive collaboration path; adjust the algorithm parameters of the ant colony algorithm when it is determined that the interaction failure rate of the target interactive collaboration path exceeds a preset failure rate threshold; and update the pheromone calculation method in the ant colony algorithm according to the target reference data.

[0147] The above-mentioned multi-agent scheduling device based on path optimization can execute the multi-agent scheduling method based on path optimization provided by any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of the execution method. For technical details not described in detail in this embodiment, please refer to the multi-agent scheduling method based on path optimization provided by any embodiment of the present invention.

[0148] Since the multi-agent scheduling device based on path optimization introduced above is a device that can execute the multi-agent scheduling method based on path optimization in the embodiment of the present invention, based on the multi-agent scheduling method based on path optimization introduced in the embodiment of the present invention, the technical personnel in this field can understand the specific implementation of the multi-agent scheduling device based on path optimization in this embodiment and its various variations, so how the multi-agent scheduling device based on path optimization implements the multi-agent scheduling method based on path optimization in the embodiment of the present invention will not be introduced in detail here. As long as the technical personnel in this field implement the device adopted by the multi-agent scheduling method based on path optimization in the embodiment of the present invention, it belongs to the scope of protection of this application.

[0149] Embodiment 4

[0150] Figure 6 A schematic diagram of the structure of an electronic device 10 that can be used to implement an embodiment of the present invention is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices (such as helmets, glasses, watches, etc.) and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present invention described and / or required herein.

[0151] like Figure 6 As shown, the electronic device 10 includes at least one processor 11, and a memory connected to the at least one processor 11, such as a read-only memory (ROM) 12, a random access memory (RAM) 13, etc., wherein the memory stores a computer program that can be executed by at least one processor, and the processor 11 can perform various appropriate actions and processes according to the computer program stored in the read-only memory (ROM) 12 or the computer program loaded from the storage unit 18 to the random access memory (RAM) 13. In the RAM 13, various programs and data required for the operation of the electronic device 10 can also be stored. The processor 11, the ROM 12, and the RAM 13 are connected to each other through a bus 14. The input / output (I / O) interface 15 is also connected to the bus 14.

[0152] A number of components in the electronic device 10 are connected to the I / O interface 15, including: an input unit 16, such as a keyboard, a mouse, etc.; an output unit 17, such as various types of displays, speakers, etc.; a storage unit 18, such as a disk, an optical disk, etc.; and a communication unit 19, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 19 allows the electronic device 10 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0153] The processor 11 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the processor 11 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any appropriate processor, controller, microcontroller, etc. The processor 11 executes the various methods and processes described above, such as a multi-agent scheduling method based on path optimization.

[0154] Optionally, a multi-agent scheduling method based on path optimization may include: obtaining team collaboration interaction data of a target team, and generating an agent interaction graph based on the team collaboration interaction data; wherein the target team includes multiple team members, and each team member represents an agent; optimizing and updating the agent interaction graph based on target reference data in the team collaboration interaction data; accumulating and updating pheromones of each interactive collaboration path in the agent interaction graph using an ant colony algorithm; selecting a target interactive collaboration path based on the pheromone concentration of each interactive collaboration path in the agent interaction graph; and instructing each of the team members to collaboratively perform team tasks based on the target interactive collaboration path.

[0155] In some embodiments, the multi-agent scheduling method based on path optimization can be implemented as a computer program, which is tangibly contained in a computer-readable storage medium, such as a storage unit 18. In some embodiments, part or all of the computer program can be loaded and / or installed on the electronic device 10 via the ROM 12 and / or the communication unit 19. When the computer program is loaded into the RAM 13 and executed by the processor 11, one or more steps of the multi-agent scheduling method based on path optimization described above can be performed. Alternatively, in other embodiments, the processor 11 can be configured to execute the multi-agent scheduling method based on path optimization in any other appropriate manner (for example, by means of firmware).

[0156] Various implementations of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), load programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various implementations can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0157] Computer programs for implementing the methods of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that when the computer program is executed by the processor, the functions / operations specified in the flow chart and / or block diagram are implemented. The computer program may be executed entirely on the machine, partially on the machine, partially on the machine and partially on a remote machine as a stand-alone software package, or entirely on a remote machine or server.

[0158] In the context of the present invention, a computer-readable storage medium may be a tangible medium that may contain or store a computer program for use by or in conjunction with an instruction execution system, device, or equipment. A computer-readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices, or equipment, or any suitable combination of the foregoing. Alternatively, a computer-readable storage medium may be a machine-readable signal medium. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0159] To provide interaction with a user, the systems and techniques described herein may be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user can provide input to the electronic device. Other types of devices may also be used to provide interaction with the user; for example, the feedback provided to the user may be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user may be received in any form (including acoustic input, voice input, or tactile input).

[0160] The systems and techniques described herein may be implemented in a computing system that includes backend components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes frontend components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such backend components, middleware components, or frontend components. The components of the system may be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), a blockchain network, and the Internet.

[0161] A computing system may include a client and a server. The client and the server are generally remote from each other and usually interact through a communication network. The client and server relationship is generated by computer programs running on the corresponding computers and having a client-server relationship with each other. The server may be a cloud server, also known as a cloud computing server or cloud host, which is a host product in the cloud computing service system to solve the defects of difficult management and weak business scalability in traditional physical hosts and VPS services.

[0162] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps recorded in this disclosure can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solutions disclosed in this disclosure can be achieved, and this document does not limit this.

[0163] The above specific implementations do not constitute a limitation on the protection scope of the present disclosure. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present disclosure shall be included in the protection scope of the present disclosure.

Claims

1. A multi-agent scheduling method based on path optimization, characterized in that: include: Acquire team collaboration interaction data of a target team, and generate an agent interaction graph according to the team collaboration interaction data; wherein the target team includes a plurality of team members, and each team member represents an agent; Optimizing and updating the agent interaction graph according to the target reference data in the team collaboration interaction data; An ant colony algorithm is used to accumulate and update the pheromones of each interactive cooperation path in the agent interaction graph; Selecting a target interactive collaboration path according to the pheromone concentration of each interactive collaboration path in the agent interaction graph; Instruct each of the team members to collaboratively perform a team task according to the target interactive collaboration path.

2. The method according to claim 1, characterized in that The team collaboration interaction data includes an agent identifier, an interaction agent object identifier, an interaction start cost, an interaction end cost, and an interaction result; Generating an agent interaction graph according to the team collaboration interaction data includes: Generate a node in the agent interaction graph according to the agent identifier and the interactive agent object identifier; Calculating a node interaction cost according to the interaction start cost and the interaction end cost; The weights of the edges between the nodes in the agent interaction graph are generated according to the node interaction costs.

3. The method according to claim 2, characterized in that The optimizing and updating the agent interaction graph according to the target reference data in the teamwork interaction data includes: Determining the type of interaction result between the nodes according to the team collaboration interaction data; Optimizing and updating the weights of the edges according to the types of interaction results between the nodes; The edges in the agent interaction graph are optimized and updated according to the types of interaction results between the nodes.

4. The method according to claim 3, characterized in that The optimizing and updating the weights of the edges according to the types of interaction results between the nodes includes: In the case where it is determined that the interaction result type is a failed interaction result, increasing the weight value of the edge corresponding to the failed interaction result; When it is determined that the interaction result type is a successful interaction result, the weight value of the edge corresponding to the successful interaction result is reduced.

5. The method according to claim 3, characterized in that: The optimizing and updating the edges in the agent interaction graph according to the types of interaction results between the nodes includes: When it is determined that the interaction result type is a failed interaction result, and when it is determined that the edge corresponding to the failed interaction result is an infeasible edge, removing the infeasible edge from the agent interaction graph; In the case where it is determined that the interaction result type is converted from a failed interaction result to a successful interaction result, edges are added between nodes that are converted from failed interaction results to successful interaction results.

6. The method according to claim 1, characterized in that The ant colony algorithm is used to accumulate and update the pheromones of each interactive cooperation path in the agent interaction graph, including: Initializing the algorithm parameters of the ant colony algorithm; wherein the algorithm parameters of the ant colony algorithm include the number of ants, pheromone volatility coefficient, pheromone importance factor, heuristic information importance factor, maximum number of iterations, and pheromone concentration of edges in the agent interaction graph; Determine a starting node and an ending node in the agent interaction graph; Placing each ant at the starting node, and each ant selects the next node in the agent interaction graph according to the state transition probability until reaching the ending node; The pheromone concentration of the edge in the agent interaction graph is updated according to the pheromone increment accumulated by each ant on the edge.

7. The method according to claim 1, characterized in that After selecting the target interactive cooperation path according to the pheromone concentration of each interactive cooperation path in the agent interaction graph, the method further includes: Calculating the interaction failure rate of the target interactive collaboration path; When it is determined that the interaction failure rate of the target interactive collaboration path exceeds a preset failure rate threshold, adjusting the algorithm parameters of the ant colony algorithm; The pheromone calculation method in the ant colony algorithm is updated according to the target reference data.

8. A multi-agent scheduling device based on path optimization, characterized in that: include: An agent interaction graph generation module is used to obtain team collaboration interaction data of a target team and generate an agent interaction graph according to the team collaboration interaction data; wherein the target team includes a plurality of team members, each of which represents an agent; An interaction graph optimization and updating module, used to optimize and update the agent interaction graph according to the target reference data in the team collaboration interaction data; A pheromone updating module, used for accumulating and updating the pheromones of each interactive cooperation path in the agent interaction graph by using an ant colony algorithm; A target interactive collaboration path selection module, used for selecting a target interactive collaboration path according to the pheromone concentration of each interactive collaboration path in the agent interaction graph; The team task execution module instructs each of the team members to collaboratively execute the team task according to the target interactive collaboration path.

9. An electronic device, characterized in that: The electronic device comprises: at least one processor; and a memory communicatively connected to the at least one processor; wherein, The memory stores a computer program executed by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute the multi-agent scheduling method based on path optimization as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a processor to implement the multi-agent scheduling method based on path optimization as described in any one of claims 1-7 when executed.

11. A computer program product comprising a computer program / instructions, wherein: When the computer program / instructions are executed by a processor, the multi-agent scheduling method based on path optimization described in any one of claims 1-7 is implemented.

Images