Adversarial Attacks on Reinforcement Learning Agents for Command and Control

arXiv:2405.01693v1 Announce Type: new
Abstract: Given the recent impact of Deep Reinforcement Learning in training agents to win complex games like StarCraft and DoTA(Defense Of The Ancients) - there has been a surge in research for exploiting learning based techniques for professional wargaming, battlefield simulation and modeling. Real time strategy games and simulators have become a valuable resource for operational planning and military research. However, recent work has shown that such learning based approaches are highly susceptible to adversarial perturbations. …

adversarial adversarial attacks agents arxiv attacks battlefield command command and control control cs.cr defense exploiting games impact modeling professional real real time research simulation strategy techniques training