07 - Systems, Subsystems and Equipment
ENHANCED SAFETY THROUGH AI: AN ASSESSMENT OF ADVERSARIAL REINFORCEMENT LEARNING FOR CONTROL OF COMPLEX AIRCRAFT
C. Koopman¹, D. Zammit-Mangion¹; ¹University of Malta, Malta
This paper explores AI integration in commercial aircraft using Reinforcement Learning and an adversarial 'Gremlin.' With a broad task space and nonrestrictive rewards, the approach aims to deepen the AI's understanding for correct response in unforeseen conditions to enhance safety. Rigorous testing uncovers insights into system robustness and limitations.