Successive Convexification for Real-Time 6-DoF Powered Descent Guidance with State-Triggered Constraints (1811.10803v1)

Abstract: In this paper, we present a real-time successive convexification algorithm for a generalized free-final-time 6-degree-of-freedom powered descent guidance problem. We build on our previous work by introducing the following contributions: (i) a free-ignition-time modification that allows the algorithm to determine the optimal engine ignition time, (ii) a tractable aerodynamics formulation that models both lift and drag, and (iii) a continuous state-triggered constraint formulation that emulates conditionally enforced constraints. In particular, contribution (iii) effectively allows constraints to be enabled or disabled by if-statements conditioned on the solution variables of the parent continuous optimization problem. To the best of our knowledge, this represents a novel formulation in the optimal control literature, and enables a number of interesting applications, including velocity-triggered angle of attack constraints and range-triggered line of sight constraints. Our algorithm converts the resulting generalized powered descent guidance problem from a non-convex free-final-time optimal control problem into a sequence of tractable convex second-order cone programming subproblems. With the aid of virtual control and trust region modifications, these subproblems are solved in succession until convergence is attained. Simulations using a third-party solver demonstrate the real-time capabilities of the proposed algorithm, with a maximum execution time of less than 0.7 seconds over a multitude of problem feature combinations.