Multi-agent cooperation method in open environment

Abstract

The invention discloses a multi-agent cooperation method in an open environment. The method comprises the following steps of S01, inputting a target task needed to be cooperatively finished; S02, automatically detecting and recording preconditions of finishing the target task in a current environment by a system; S03, according to the input target task and the preconditions in the current environment, looking for agents capable of finishing the task in a current ability base; if the agents meeting the requirements are not found, going to the step S04; if the agents capable of finishing the task exist, going to the step S05; S04, suspending execution of the current task, and after a new agent capable of finishing the task is added, going to the step S02 and continuing to execute the step S02; S05, executing the task needed to be finished by the agents, and after the execution is finished, going to the step S02 and continuing to execute the step S02, until all tasks are finished; and S06, generating a corresponding commitment protocol according to an executive sequence, storing the commitment protocol in the ability base, and if similar tasks also exist the next time, performing direct calling. According to the method, multiple agents can adopt an ability matching algorithm to finish the target task according to change of a dynamic environment and the corresponding commitment protocol is generated.

Description

technical field [0001] The invention relates to the technical field of multi-agent cooperation, in particular to a multi-agent cooperation method in an open environment. Background technique [0002] Agent collaboration technology is an important part of the research on multi-agent systems in the field of artificial intelligence. When a single agent cannot complete a task, multiple agents need to cooperate to complete a certain task. The entire collaboration process is usually controlled by a commitment protocol. [0003] Usually, the commitment agreement is not dynamically generated, but defined before executing the task and embedded in the process of multi-agent executing the task. However, this way of pre-defining commitment agreements has many limitations and deficiencies: [0004] (1) When the agent changes [0005] Due to the openness of the system, new agents will enter the system at any time, and the predefined commitment protocol often cannot meet the requirement...

AI Technical Summary

Problems solved by technology

However, this way of pre-defining commitment agreements has many limitations and deficiencies:

[0015] Method 1 needs to provide a commitment agreement in advance, and then modify the commitment agreement according to different environments or tasks. It cannot automatically modify the commitment agreement according to changes in the dynamic environment, which has poor adaptability and heavy workload.

Method 2 and method 3 have a large amount of data processing, cumbersome calculations and need to occupy a powerful system memory

Images