Independent Soft Actor-Critic Deep Reinforcement Learning for UAV Cooperative Air Combat Maneuvering Decision-Making

Abstract

This paper delves into the research of collaborative combat strategies for multiple unmanned combat aerial vehicles (UAVs), utilizing the independent soft Actor-Critic (is-AC) algorithm. We aim to achieve collaborative jamming confrontation, accurate battlefield situational awareness, and UAV decision-making capabilities to control their behavior. However, the SAC algorithm is plagued by instability and poor scalability in Multi-agent reinforcement learning scenarios. To address this, we draw inspiration from the Independent Q-Learning (IQL) algorithm and improve SAC. Our experimental analysis of the is-AC algorithm in UAV confrontation models demonstrates its stability and scalability in multi-machine scenarios.