Autonomous Flight, Fault, and Energy Management of the Flying Fish Solar-Powered Seaplane.

Abstract

The Flying Fish autonomous unmanned seaplane is designed and built for persistent ocean surveillance. Solar energy harvesting and always-on autonomous control and guidance are required to achieve unattended long-term operation. This thesis describes the Flying Fish avionics and software systems that enable the system to plan, self-initiate, and autonomously execute drift-flight cycles necessary to maintain a designated watch circle subject to environmentally influenced drift. We first present the avionics and flight software architecture developed for the unique challenges of an autonomous energy-harvesting seaplane requiring the system to be: waterproof, robust over a variety of sea states, and lightweight for flight. Seaplane kinematics and dynamics are developed based on conventional aircraft and watercraft and upon empirical flight test data. These models serve as the basis for development of flight control and guidance strategies which take the form of a cyclic multi-mode guidance protocol that smoothly transitions between nested gain-scheduled proportional-derivative feedback control laws tuned for the trim conditions of each flight mode. A fault-tolerant airspeed sensing system is developed in response to elevated failure rates arising from pitot probe water ingestion in the test environment. The fault-tolerance strategy utilizes sensor characteristics and signal energy to combine redundant sensor measurements in a weighted voting strategy, handling repeated failures, sensor recovery, non-homogenous sensors, and periods of complete sensing failure. Finally, a graph-based mission planner combines models of global solar energy, local ocean-currents, and wind with flight-verified/derived aircraft models to provide an energy-aware flight planning tool. An NP-hard asymmetric multi-visit traveling salesman planning problem is posed that integrates vehicle performance and environment models using energy as the primary cost metric. A novel A* search heuristic is presented to improve search efficiency relative to uniform cost search. A series of cases studies are conducted with surface and airborne goals for various times of day and for multi-day scenarios. Energy-optimal solutions are identified except in cases where energy harvesting produces multiple comparable-cost plans via negative-cost cycles. The always-on cyclic guidance/control system, airspeed sensor fault management algorithm, and the nested-TSP heuristic for A* are all critical innovation required to solve the posed research challenges.