Multi-agent action strategy learning method and device, medium and computing equipment

Abstract

The embodiment of the invention provides a multi-agent action strategy learning method. The multi-agent action strategy learning method comprises the steps that multiple agents sample corresponding actions according to respective initial action strategies; respectively estimating the advantages obtained after the multiple agents execute the corresponding actions; and updating the action strategy of each intelligent agent based on the advantages obtained after the multiple intelligent agents execute the corresponding actions, so that each updated action strategy can enable the corresponding intelligent agent to obtain higher return. The method provided by the invention is applied to a task processing-oriented machine learning scene; meanwhile, a plurality of cooperative intelligent agents are trained (namely a plurality of action strategies are trained at the same time). A pre-built simulator and the intelligent agents are not adopted for interaction, manual supervision is not needed, time cost and resources are greatly saved, in addition, in order to enable all the intelligent agents to learn excellent action strategies, different awards are distributed to all the intelligent agents, and therefore the multiple intelligent agents can learn the more excellent action strategies.

Description

technical field [0001] Embodiments of the present invention relate to the field of reinforcement learning, and more specifically, embodiments of the present invention relate to a multi-agent action strategy learning method, device, medium, and computing device. Background technique [0002] This section is intended to provide a background or context for implementations of the invention that are recited in the claims. The descriptions herein are not admitted to be prior art by inclusion in this section. [0003] The action policy determines the next action the agent should take and plays a crucial role in task-oriented systems. In recent years, policy learning has been widely regarded as a reinforcement learning (RL) problem. Since RL requires a lot of interaction for policy training, it is time-consuming and labor-intensive to interact directly with real users. The most common approach is to develop a user simulator to help train the target agent to learn action strategie...

AI Technical Summary

Problems solved by technology

[0004] However, designing a reliable user simulator is not trivial and often challenging, since it is synonymous with building a good agent

Images