Trajectory optimization with detection avoidance for visually identifying an aircraft

Abstract

Unmanned aerial vehicles (UAVs) play an essential role for the US Armed Forces by performing missions deemed as "dull, dirty and dangerous" for a pilot. As the capability of UAVs expand. they will perform a broader range of missions such as air-to-air combat. The focus of this thesis is forming trajectories for the closing phase of an air-to-air combat scenario. A UAV should close with the suspected aircraft in a manner that allows a ground operator to visually identify the suspected aircraft while avoiding visual/electronic detection from the other pilot. This thesis applies and compares three methods for producing trajectories which enable a visual identification. The first approach is formulated as a mixed integer linear programming problem which can be solved in real time. However, there are limitations to the accuracy of a radar detection model formed with only linear equations, which might justify using a nonlinear programming formulation. With this approach the interceptor's radar cross section and range between the suspected aircraft and interceptor can be incorporated into the problem formulation. The main limitation of this method is that the optimization software might not be able to reach online an optimal or even feasible solution. The third applied method is trajectory interpolation. In this approach, trajectories with specified boundary values and dynamics are formed offline; online, the method interpolates between the given trajectories to obtain similar maneuvers with different initial conditions and end- states. With this method, because the number of calculations required to produce a feasible trajectory is known, the amount of time to calculate a trajectory can be estimated.