Research on UAV Swarm Confrontation Task Based on MADDPG Algorithm
Lei Xiang, Tao Xie
Abstract
In recent years, with the rapid development of UAV technology, the demand of anti-UAV technology is increasingly urgent. Among the existing soft and hard means to counter the attack of UAV swarm, countering UAV swarm by UAV swarm is an important way to counter the attack of UAV swarm in the future. Based on the idea of UAVs intelligent attack and defense confrontation, this paper establishes a simulation environment of UAVs confrontation and a intelligent model of UAV swarm based on MADDPG algorithm. Aiming at the problems such as the speed control of UAV is not accurate and it is difficult to choose the appropriate attack angle in the confrontation, a rule-coupled method is proposed to effectively improve the confront ability of UAV. The experimental results show that the rule-coupled method can significantly improve the winning rate of the UAV swarm in the confrontation from 63% to 81%, and reduce the average epochs required to destroy all the enemy UAVs in a winning game from 61 to 49.