Author(s)

Qianying Li, Baolong Guo, Zhe Huang and Cheng Li

Corresponding Author

Qianying Li

ABSTRACT

The development of machine learning technology has so far no longer relied entirely on artificial power in adversarial decision-making, but there has been an updated decision competition. The climax of the development of artificial intelligence and machine computing capabilities has provided powerful technical support and hardware support for the development of intelligent adversarial systems. The improved actor-critic extended multi-agent strategy gradient method (ACEM) is mainly used to optimize the behavior decision of multi-agent cooperation and hostility. This algorithm introduces extended strategy review information on the basic actor-critic framework for reward calculation. The use of adversarial strategy inference reduces the impact of environmental instability on strategy selection. This method has significantly improved the degree of completion of target tasks in cooperative and hostile environments.

KEYWORDS

Multi-agent; Adversarial System; Deep Reinforcement Learning; Actor-Critic Framework