Self-Designing Control Systems for Piloted and Uninhabited Aerial Vehicles

摘要

The main advantage of self-designing flight control systems is their ability to optimize performance automatically, resulting in a substantial reduction of the cost and time needed for control law development. Self- designing control systems also reconfigure automatically after failures and damages, yielding greater chances of survival in dangerous conditions. A self- designing nonlinear autopilot was developed to interface with the high-level path planning of a UAV. The control algorithm is distinct from conventional autopilots in that it is not based on a known, linearized model of the aircraft. Instead, the algorithm compensates for nonlinear dynamic effects and adjusts its parameters automatically, exploiting the reconfiguration capabilities of an inner control loop designed using adaptive methods. A new control allocation algorithm was also developed, based on the direct allocation method of Durham. A special representation using spherical coordinates was used to speed-up the computations that must be performed at a high sampling rate. The direct allocation method was also extended to a class of systems that had previously been excluded, namely those for which some independent control surfaces produce linearly dependent moments. Finally, fast algorithms for optimal control allocation were developed based on linear programming techniques. It was observed that significant improvements in performance could be obtained at the cost of only modest increases in computations.