EA2DM: Energy-Aware Decision Making for Autonomous UAS

Autonomous operations of an Unmanned Aerial System (UAS) and complex mission are governed by safety and resource consumption. All missions (military and commerical) must be performed safely, even in the case of system or component failure or unexpected environmental conditions. Under such circumstances, as many mission objectives as possible must be carried out successfully. On-board monitoring and contingency management can provide these capabilities. Our proposed work, EA2DM (Energy-Aware Autonomous Decision Making), will enable certifiable, on-board, real-time energy-aware contingency management. During flight, the on-board EA2DM attempts to sustain overall mission performance of the UAS in case of failures or unexpected environmental conditions by balancing energy optimization, mission objectives, and safety. EA2DM will substantially increase safety for the UAS operation and enable successful and prolonged missions even in adverse cases. The EA2DM architecture combines algorithms for advanced anomaly detection, model-based diagnosis, and prognostics with a hybrid model-based decision engine. During flight, anomaly detection and diagnosis will continuously provide up-to-date health information of the UAS. In case of failures or unexpected environmental conditions, EA2DM will trigger its on-board decision-making to produce a mitigation plan that fulfills safety constraints, optimize resource and energy usage, and enables the UAS to meet as many mission objectives as are still possible. EA2DM uses a hybrid symbolic-numerical approach for this high-dimensional optimization problem. The project team will define 3 challenging and realistic Use Cases to evaluate our approach.