Explaining AI decisions in autonomous satellite scheduling via computational argumentation

The task of scheduling satellite operations is inherently complex and highly sensitive to alterations, a challenge compounded by the increasing number of satellites in orbit. The escalating risks and complexities have prompted organizations to explore automated solutions to replace traditional manual processes. However, concerns about the trustworthiness and transparency of automated systems prevent their widespread adoption.

eXplainable Artificial Intelligence (XAI) is an emerging field that aims to address these reservations by enabling Artificial Intelligence (AI) systems to provide explanations for their decisions, thereby eliminating opaqueness in understanding their reasoning. Within XAI, the use of computational argumentation frameworks has seen increasing utilisation. This approach quantifies the supportability of decisions, offering system operators enhanced understanding and justification for utilizing automated services.

This paper expands on previous research by detailing a method for generating a tripolar argumentation approach for assessing actions based on an Earth Observation (EO) satellite schedule. The method involves calculating and presenting the weights of arguments that support or attack the scheduled actions. The results illustrate the effectiveness of the approach in producing meaningful insights into scheduling decisions, highlighting its potential for practical applications in real-world satellite operations.