CN-120773027-B - Multi-robot cooperation method and system under dynamic environment and electronic equipment

CN120773027BCN 120773027 BCN120773027 BCN 120773027BCN-120773027-B

Abstract

The embodiment of the invention provides a multi-robot cooperation method, a multi-robot cooperation system and electronic equipment in a dynamic environment, which relate to the technical field of robots and comprise the steps of generating a task relation dependency graph represented by a Directed Acyclic Graph (DAG) corresponding to a task by utilizing task description information, generating a layered DAG corresponding to each subtask through a large language model LLMs based on first environment description information of the current environment, functional information of each robot and the DAG corresponding to the task, wherein the layered DAG represents an execution strategy of the subtask, obtaining a task plan of each robot based on the functional information and first state information of each robot and the subtask execution strategy corresponding to each subtask, and distributing the task plan of each robot to the robots so as to enable each robot to execute the task plan. Therefore, the flexibility of task planning and distribution in the multi-robot cooperation process is improved, and the method is better suitable for dynamic environments.

Inventors

GUO MENG
WEI JINSHENG
CHEN JUNFENG
LI ZHONGKUI

Assignees

北京大学

Dates

Publication Date: 20260505
Application Date: 20250625

Claims (9)

1. A multi-robot collaboration method in a dynamic environment, comprising: Receiving task description information; Generating a task relation dependency graph represented by a Directed Acyclic Graph (DAG) corresponding to a task by utilizing the task description information, wherein each node in the DAG represents a subtask, and edges between every two nodes represent the dependency relation between the subtasks represented by the two nodes; Inputting the first environment description information of the current environment, the function information of each robot and the DAG corresponding to the task into a large language model LLMs, and outputting a first execution strategy through the large language model LLMs; Acquiring a history execution strategy of a history task with the same task type as that of a current task, inputting the history execution strategy into the large language model LLMs, and adjusting the first execution strategy through the large language model LLMs to obtain a second execution strategy; Acquiring a resource conflict constraint relation and a time conflict constraint relation, inputting the resource conflict constraint relation and the time conflict constraint relation into the large language model LLMs, and adjusting the second execution strategy through the large language model LLMs to obtain a hierarchical DAG corresponding to each subtask, wherein the hierarchical DAG represents the execution strategy of the subtask; acquiring a task plan of each robot based on the function information and the first state information of each robot and the subtask execution strategy corresponding to each subtask; the mission plan of each robot is assigned to the robot such that each robot executes the mission plan.
2. The method of claim 1, wherein when the task description information includes a task description in a natural language representation, the method further comprises, after the receiving task description information: And converting the task description of the natural language representation into task description information in the form of linear sequential logic LTL.
3. The method according to claim 1, wherein the obtaining the task plan of each robot based on the function information and the first state information of each robot and the subtask execution policy corresponding to each subtask includes: based on the function information and the first state information of each robot and the subtask execution strategy corresponding to each subtask, the task plan of each robot is obtained by taking the minimum overall task completion time as an optimization target through a branch-and-bound search method and an integer programming method.
4. The method of claim 1, wherein after said assigning a mission plan for each robot to said robot, said method further comprises: When a first preset condition is met, acquiring second environment description information, wherein the first preset condition comprises at least one of detecting a new operation object corresponding to a task type of a current task in a current environment, detecting a new resource type in the current environment, detecting a new resource of an existing resource type in the current environment and detecting a new task type, and the second environment description information is updated environment description information compared with the first environment description information; And regenerating the hierarchical DAG corresponding to each subtask through a large language model LLMs based on the second environment description information, the function information of each robot and the updated DAG.
5. The method of claim 1, wherein after said assigning a mission plan for each robot to said robot, said method further comprises: when a second preset condition is met, second state information of each robot is obtained, wherein the second preset condition comprises that the plan completion information fed back by the robot for executing the task plan is not received in a preset time range, or that the fault of the robot is detected, and the second state information is updated compared with the first state information; and obtaining updated task plans of the robots based on the function information and the second state information of the robots and the subtask execution strategy corresponding to the subtasks.
6. The method of claim 1, wherein after generating the task relationship dependency graph represented by the task-corresponding directed acyclic graph, DAG, the method further comprises exposing the DAG to cause a human operator to verify the DAG; After the generating the hierarchical DAG corresponding to each subtask, the method further comprises displaying the hierarchical DAG corresponding to each subtask so that a human operator verifies the hierarchical DAG corresponding to each subtask; After the obtaining the mission plan of each robot, the method further includes displaying the mission plan of each robot to enable a human operator to verify the mission plan of each robot; after the task plans of each robot are distributed to the robots, the method further comprises the steps of receiving plan execution progress information fed back by the robots and displaying the plan execution progress information so that a human operator verifies the task plans of the robots based on the plan execution progress information.
7. The method according to any one of claims 1 to 6, wherein the first environment description information includes task object information of an operation object corresponding to a task type of a current task and resource information of resources in an environment, and the function information of the robot indicates an operation type that the robot can perform and a resource type that can operate.
8. The multi-robot cooperation system in the dynamic environment is characterized by comprising a task understanding module, a subtask generating module and a subtask distributing module; The task understanding module is used for receiving task description information, generating a task relation dependency graph represented by a Directed Acyclic Graph (DAG) corresponding to a task by utilizing the task description information, wherein each node in the DAG represents a subtask, and the edge between every two nodes represents the dependency relation between the subtasks represented by the two nodes; The subtask generation module is used for inputting first environment description information of a current environment, functional information of each robot and task corresponding DAG into a large language model LLMs, outputting a first execution strategy through the large language model LLMs, acquiring a history execution strategy of a history task with the same task type as the current task, inputting the history execution strategy into the large language model LLMs, adjusting the first execution strategy through the large language model LLMs to obtain a second execution strategy, acquiring a resource conflict constraint relation and a time conflict constraint relation, inputting the resource conflict constraint relation and the time conflict constraint relation into the large language model LLMs, adjusting the second execution strategy through the large language model LLMs to obtain layered DAGs corresponding to each subtask, wherein the layered DAGs represent the execution strategies of the subtasks; The subtask allocation module is used for obtaining a task plan of each robot based on the function information and the first state information of each robot and the subtask execution strategy corresponding to each subtask, and allocating the task plan of each robot to the robots so that each robot executes the task plan.
9. The electronic equipment is characterized by comprising a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory are communicated with each other through the communication bus; a memory for storing a computer program; a processor for implementing the method of any of claims 1-7 when executing a program stored on a memory.

Description

Multi-robot cooperation method and system under dynamic environment and electronic equipment Technical Field The present invention relates to the field of robots, and in particular, to a method, a system, and an electronic device for multi-robot collaboration in a dynamic environment. Background In the multi-robot cooperation process, task planning and distribution of multiple robots are important processes. In the field of robot task planning and distribution, tasks generally need to be broken down into a series of sub-tasks that must be able to be performed by the deployed robot. In the related art, task decomposition is generally implemented by classical methods, such as manual planning methods, in which task decomposition logic is defined by manually set rules, and planning and allocation policies are generally constructed manually according to domain expertise and hard-coded. The method is simple and visual, but has poor flexibility, and is difficult to adapt to dynamic environments. Disclosure of Invention The embodiment of the invention aims to provide a multi-robot cooperation method, a system and electronic equipment in a dynamic environment, so as to improve the flexibility of task planning and allocation in the multi-robot cooperation process and better adapt to the dynamic environment. The specific technical scheme is as follows: In a first aspect, a multi-robot collaboration method in a dynamic environment is provided, including: Receiving task description information; Generating a task relation dependency graph represented by a Directed Acyclic Graph (DAG) corresponding to a task by utilizing the task description information, wherein each node in the DAG represents a subtask, and edges between every two nodes represent the dependency relation between the subtasks represented by the two nodes; Generating a hierarchical DAG corresponding to each subtask through a large language model LLMs based on the first environment description information of the current environment, the function information of each robot and the DAG corresponding to the task, wherein the hierarchical DAG represents an execution strategy of the subtask; acquiring a task plan of each robot based on the function information and the first state information of each robot and the subtask execution strategy corresponding to each subtask; the mission plan of each robot is assigned to the robot such that each robot executes the mission plan. Optionally, when the task description information includes a task description expressed in natural language, after the receiving the task description information, the method further includes: And converting the task description of the natural language representation into task description information in the form of linear sequential logic LTL. Optionally, the generating, based on the first environment description information, the function information of each robot, and the task-corresponding DAG, a hierarchical DAG corresponding to each subtask through a large language model LLMs includes: inputting the first environment description information, the function information of each robot and the task-corresponding DAG into a large language model LLMs, and outputting a first execution strategy through the large language model LLMs; Acquiring a history execution strategy of a history task with the same task type as that of a current task, inputting the history execution strategy into the large language model LLMs, and adjusting the first execution strategy through the large language model LLMs to obtain a second execution strategy; And acquiring a resource conflict constraint relation and a time conflict constraint relation, inputting the resource conflict constraint relation and the time conflict constraint relation into the large language model LLMs, and adjusting the second execution strategy through the large language model LLMs to obtain a hierarchical DAG corresponding to each subtask. Optionally, the obtaining a task plan of each robot based on the function information and the first state information of each robot and the subtask execution policy corresponding to each subtask includes: based on the function information and the first state information of each robot and the subtask execution strategy corresponding to each subtask, the task plan of each robot is obtained by taking the minimum overall task completion time as an optimization target through a branch-and-bound search method and an integer programming method. Optionally, after the assigning of the mission plan for each robot to the robot, the method further comprises: When a first preset condition is met, acquiring second environment description information, wherein the first preset condition comprises at least one of detecting a new operation object corresponding to a task type of a current task in a current environment, detecting a new resource type in the current environment, detecting a new resource of an existing resource type in th