Explainable AI in Aerospace for Enhanced System Performance
Sujitra Sutthithatip, Suresh Perinpanayagam, Sohaib Aslam, Andrew Wileman
Abstract
Explainable AI (XAI) is maturing as a pivotal subfield of AI, focused on uncovering complex AI models to human users in a highly systematic, interpretable, and understandable manner. This paper is a snapshot of XAI implementation in aerospace, reflecting the existing discourse in this field with respect to decision-making in critical situations. Supported by a comprehensive XAI taxonomy, the explanation framework is put forth with human-centric and scientific explanations as its integral components to help describe and interpret the model outcome. A use-case formed around the aircraft operation and its essential constituents - the pilots, air-traffic controllers, maintainers, and aircraft manufacturers, is presented as a representative implementation of XAI. It is shown how the decisions that come along with the meaningful information can help operators such as pilots and air traffic controllers’ decisions, so they can consider information consciously to make decisions within a certain time.