Sampling-Based Planning for Guaranteed Safe Energy Management of Hybrid UAV Powertrain Under Complex, Uncertain Constraints

As electrified aircraft are becoming more prominent, new energy management strategies are needed to fully leverage their capabilities to perform more complex missions and to do so safely. Nonconvex constraints, multitimescale dynamics, and uncertainty introduce challenges in the way of guaranteeing safe powertrain operation using existing methods. This work seeks to address these challenges using a novel application of sampling-based planning methods to plan the operation of a hybrid unmanned aerial vehicle (UAV) powertrain. Known for their computational efficiency, these sampling-based methods can rapidly react to changing mission information. A two-stage method is introduced, which manages multiple time scales using rapidly exploring random tree (RRT)-based algorithms for long-term planning and robust model predictive control (RMPC) for short-term execution of mission plans with guaranteed tracking error bounds. An experimentally validated case study demonstrates the implementation of the two-stage method using RRT-based algorithms. Rapid planning times (
$>$
100
$\times$
faster than real time) enable replanning online to react to changing mission specifications. Robust tracking control guarantees that the UAV powertrain is safely operated in the presence of complex, uncertain constraints.