Verification of successive convexification algorithm
Berning, Andrew Walter, Jr.
MetadataShow full item record
In this report, I describe a technique which allows a non-convex optimal control problem to be expressed and solved in a convex manner. I then verify the resulting solution to ensure its physical feasibility and its optimality. The original, non-convex problem is the fuel-optimal powered landing problem with aerodynamic drag. The non-convexities present in this problem include mass depletion dynamics, aerodynamic drag, and free final time. Through the use of lossless convexification and successive convexification, this problem can be formulated as a series of iteratively solved convex problems that requires only a guess of a final time of flight. The solution’s physical feasibility is verified through a nonlinear simulation built in Simulink, while its optimality is verified through the general nonlinear optimal control software GPOPS-II.