Model-based on-board decision making for autonomous aircraft

Abstract

Powerful, small and lightweight sensors in combination with advanced failure detection, diagnosis, and prognostics techniques provide up-to-date data on the health status of an Unmanned Aerial System (UAS) or autonomously piloted vehicle. This information must be used for automatic planning and execution of contingency actions to keep the UAS safe in adverse conditions. We present DM (Decision Maker), a software component which uses model-based reasoning and backtracking search to iteratively construct contingency plans that are safe for the UAS to execute and pose minimal interruption to the mission goals. The DM is a discrete decision making system has been developed within the NASA Autonomous Operating System (AOS) project and fills the gap between Prognostics and Health Management and autonomous flight operations. In this paper, we describe DM and its reasoning algorithm and present the supporting modeling framework for the construction of system and fault models. Flights with a DJI S1000+ octocopter with fault injection will be used as our case study.