V&V of Automated Mission Planning for Unmanned Rotorcraft

Abstract

The DLR Autonomous Research Test-bed for Intelligent Systems platform (ARTIS) comprises a fleet of unmanned research rotorcraft that is used to develop functions for autonomous flight. The mission planning function is an essential part of our research efforts for autonomous mission execution. From high-level task specifications it generates flight trajectories and actions in known and unknown environments. The software decides about constrained sequences of actions and how to avoid obstacles. Automated decision making during flight requires a high degree of software trustworthiness through extensive software verification efforts. But despite all efforts, validation and verification problems arise when trying to fulfill safety standards for UAVs. This is especially true for software standards, such as DO-178B/C. There are multiple challenges: Large workforce requirements, Non-finalized regulations and DAL level requirements for software component certification, High runtime complexity for verification of autonomous behaviour, Balancing efforts for experimental autonomous functions, e.g. for research purposes. Therefore, we discuss our ARTIS validation and verification efforts for a trustworthy decision and control software in order to assure a certain level of DO-178C compliance. The V&V techniques for our automated mission planner are summarized, and a short outlook in advanced methods is provided which may improve verification results and efforts. The test concept also assesses the performance (i.e. approach quality) and thus exceeds assessment of sole functional correctness. This is done using criteria like path derivation, safety distance or real-time limits. Furthermore, our “best practices” are reviewed with respect to the DO-178C standard.