An Abstract Architecture for Explainable Autonomy in Hazardous Environments