Real-time trajectory optimization of UAV using sequential convex programming

Abstract

In this thesis, sequential convex programming is applied to real-time trajectory optimization of Unmanned Aerial Vehicle(UAV). In the existing trajectory optimization, there have been many studies for generating guidance commands based on the analytic solution, but there is a problem that it does not reflect various constraints of UAV in actual operation. Recently, however, the computational guidance method has been developed to directly calculate the optimization problem numerically due to the development of computer processor performance. One of the various computational guidance methods, the second-order cone programming(SOCP) is known to be able to solve quickly by an interior point method using the characteristics of convex optimization that converges to the global optimal solution at all times. Therefore, sequential convex programming that approximates a non-convex problem sequentially by the SOCP method has appeared. Using these features, it is shown that real-time trajectory optimization is possible by assuming various aerial mission scenarios of UAV. Also, an algorithm is introduced to avoid various spherical and polyhedral obstacles, which is nonconvex, into a convex programming method. Then the proposed method is compared with the pseudospectral method, which is a conventional optimization method and shows the advantage in calculation time.