CN-121979682-A - Intelligent agent task execution method, system, medium and electronic equipment

Abstract

The embodiment of the disclosure relates to the technical field of intelligent assistants, and provides an agent task execution method, an agent task execution system, an agent task execution medium and an electronic device. The method comprises the steps that a first end sends a target task to a second end, the second end starts from a unified starting point state of the target task, a plurality of branch intelligent agents are utilized for parallel exploration to obtain a target path corresponding to the target task, the second end sends the target path to the first end, and the first end utilizes a main intelligent agent to execute the target task based on the target path. According to the method provided by the disclosure, the target path is obtained by utilizing the branch intelligent agent to perform path exploration at the second end, and the target task is executed by the main intelligent agent at the first end, so that the task execution success rate of the main intelligent agent can be improved.

Inventors

• LIN WAI
• XUE ZIZHEN
• GONG YICHEN

Assignees

• XG科技私人有限公司

Dates

Publication Date

20260505

Application Date

20260127

Claims (14)

1. 1. An agent task execution method, the method comprising: the first end sends the target task to the second end; The second end starts from the unified starting point state of the target task, and a plurality of branch intelligent agents are utilized for parallel exploration to obtain a target path corresponding to the target task; the second end sends the target path to the first end; The first end executes the target task based on the target path using a primary agent.
2. 2. The agent task execution method according to claim 1, wherein the second end starts from a unified starting point state of the target task, and obtains a target path corresponding to the target task by using a plurality of branch agents to search in parallel, and the method comprises: the second end determines at least one exploration phase based on the target task; The second end determines a unified stage starting point state corresponding to the first exploration stage based on the unified starting point state of the target task; for each exploration stage, the second end starts from a unified stage starting point state corresponding to the exploration stage, and utilizes a plurality of branch intelligent agents to conduct parallel exploration to obtain a target path segment corresponding to the exploration stage, and determines the unified stage starting point state corresponding to the next exploration stage based on an end point state reached by the target path segment; and the second end group determines a target path corresponding to the target task based on the target path segments corresponding to the exploration phases.
3. 3. The agent task execution method according to claim 2, wherein the second end determines a target path corresponding to the target task based on a target path segment corresponding to each exploration phase, including: the second end group judges whether the finishing condition corresponding to the target task is reached or not based on the end state reached by the target path segment corresponding to the last exploration stage of the sequencing; And in response to the completion condition, the second end splices target path segments corresponding to each exploration stage to obtain a target path corresponding to the target task.
4. 4. The method for executing an agent task according to claim 2, wherein the second end starts from a unified stage start state corresponding to the exploration stage, and obtains the target path segment corresponding to the exploration stage by using a plurality of branch agents to explore in parallel, and the method comprises: The second end starts from the unified stage starting point state by utilizing the plurality of branch intelligent agents, and executes the target task in parallel in a virtual environment set to the unified stage starting point state to obtain candidate path segments corresponding to the branch intelligent agents; and the second end evaluates each candidate path segment and determines the target path segment from a plurality of candidate path segments according to the evaluation result.
5. 5. The agent task execution method according to any one of claims 2 to 4, wherein the second end performs parallel exploration using a plurality of branch agents starting from a unified stage start state corresponding to the exploration stage, including: the second end starts from a unified stage starting point state corresponding to the exploration stage, and utilizes the plurality of branch agents to explore in parallel based on exploration strategies corresponding to the branch agents.
6. 6. The agent task execution method according to claim 4, wherein the executing the target task in parallel, to obtain candidate path segments corresponding to the branch agents, includes: Based on the target task and the current state information of the virtual environment, iteratively executing at least one operation step in parallel until the branch intelligent agent reaches a termination condition corresponding to the exploration phase; The second end group determines candidate path segments corresponding to each of the branch agents based on the at least one operation step performed by each of the branch agents.
7. 7. The agent task execution method according to claim 6, wherein the branch agent reaching a termination condition corresponding to the exploration phase includes: The branch agent performs a preset number of the operating steps in the exploration phase or, And the branch agent executes the target task in the exploration stage for a preset duration.
8. 8. The agent task execution method according to any one of claims 1 to 4, wherein the second end starts from a unified starting point state of the target task, and before parallel exploration with a plurality of branch agents, the method further comprises: the first end sends current equipment state information to the second end; the second end determines the unified starting point state of the target task based on the current equipment state information.
9. 9. The agent task performing method of claim 8, wherein the first end sends current device state information to the second end, comprising: capturing screen capturing information and interface layout information in the current equipment state information by the first end; the first end sends the screen capturing information and the interface layout information to the second end.
10. 10. The agent task execution method of claim 4, wherein the second end starts from the unified phase start state with the plurality of branch agents, and before executing the target task in parallel in the virtual environment set to the unified phase start state, the method further comprises: and the second end synchronizes the virtual environment where each branch intelligent agent is located to a unified stage starting point state corresponding to the exploration stage.
11. 11. The agent task execution method of claim 4, wherein the second end starts from a unified stage start state corresponding to the exploration stage, and before parallel exploration with a plurality of branch agents, the method further comprises: The first end sends the equipment operation environment information corresponding to the first end to a second end; The second end establishes a plurality of virtual environments matched with the equipment operation environment of the first end based on the equipment operation environment information, and deploys at least one branch intelligent agent in each virtual environment.
12. 12. An agent task execution system, the system comprising: The main intelligent body is deployed in the first end and is used for sending the target task to the second end; the branch intelligent agents are deployed in the second end and used for exploring in parallel from the unified starting point state of the target task to obtain a target path corresponding to the target task; A sending unit, configured to be disposed in the second end, and configured to send the target path to the first end; the main agent is further configured to execute the target task based on the target path.
13. 13. A computer readable storage medium having stored thereon computer program instructions which, when executed by a processor, cause the processor to perform the method of any of claims 1 to 11.
14. 14. An electronic device comprising one or more processors and memory, the memory storing computer instructions that, when executed by the processor, cause the processor to perform the method of any one of claims 1 to 11.

Description

Intelligent agent task execution method, system, medium and electronic equipment Technical Field The disclosure relates to the technical field of artificial intelligence, and in particular relates to an agent task execution method, an agent task execution system, an agent task execution medium and an electronic device. Background A Graphic User Interface (GUI) agent (hereinafter referred to as agent) is an artificial intelligence program capable of simulating human interaction with a software interface, and performing operations (e.g., clicking, sliding, inputting text) on various software interfaces to accomplish a specific task, thereby improving operation efficiency and automation level. Currently, the agent can be applied to scenes needing to automatically operate a graphical user interface, such as scenes of Robot Process Automation (RPA), automated software testing, intelligent personal assistant and the like. Disclosure of Invention When the intelligent agent executes the task, the intelligent agent can only make a decision based on the current real-time information, and once an error path is selected at a certain decision point, the task execution can fail. In order to solve the above technical problems, embodiments of the present disclosure provide an agent task execution method, system, medium, and electronic device, where a target path is obtained by performing path exploration by using a branch agent at a second end, and a target task is executed by a main agent at a first end, so that a task execution success rate of the main agent can be improved. The first aspect of the disclosure provides an agent task execution method, which comprises the steps that a first end sends a target task to a second end, the second end starts from a unified starting point state of the target task, a target path corresponding to the target task is obtained through parallel exploration of a plurality of branch agents, the second end sends the target path to the first end, and the first end executes the target task based on the target path by using a main agent. The second aspect of the disclosure provides an agent task execution system, which comprises a main agent deployed in a first end and used for sending a target task to a second end, a plurality of branch agents deployed in the second end and used for exploring in parallel from a unified starting point state of the target task to obtain a target path corresponding to the target task, a sending unit deployed in the second end and used for sending the target path to the first end, and the main agent is further used for executing the target task based on the target path. In a third aspect of the present disclosure, embodiments of the present disclosure provide a computer-readable storage medium storing a computer program for executing the agent task execution method provided in the first aspect. In a fourth aspect of the present disclosure, an embodiment of the present disclosure provides an electronic device, where the electronic device includes a processor, and a memory configured to store executable instructions of the processor, where the processor is configured to read the executable instructions from the memory and execute the executable instructions to implement the method for executing an agent task provided in the first aspect. According to the method for executing the task of the agent, when the first end needs to execute the target task by using the main agent, the second end is firstly searched by the plurality of branch agents in parallel to obtain a target path, and then the main agent at the first end executes the target task based on the target path. Because the target path is obtained by the parallel exploration of the branch intelligent agents at the second end, the risk of single-point decision errors of the intelligent agents can be reduced, and the success rate of task execution is improved. On the basis, the second end is explored in parallel by a plurality of branch intelligent agents, and the possible execution paths explored by the plurality of branch intelligent agents can be evaluated and preferential, so that the balance problem between exploration (trying unknown operation sequences) and utilization (using known effective paths) is solved, and a better solution is obtained as a target path. Drawings The above and other objects, features and advantages of the present disclosure will become more apparent by describing embodiments thereof in more detail with reference to the accompanying drawings. The accompanying drawings are included to provide a further understanding of embodiments of the disclosure, and are incorporated in and constitute a part of this specification, illustrate embodiments of the disclosure and together with the description serve to explain the disclosure, without limitation to the disclosure. In the drawings, like reference numerals generally refer to like parts or steps. Fig. 1 is a schematic diagram of an agent ta