Implementation and Experimental Demonstration of Onboard Powered-Descent Guidance

Onboard, fuel-optimal, constrained powered-descent guidance based on the theory of lossless convexification has been implemented as the Guidance for Fuel-Optimal Large Diverts (G-FOLD) algorithm. Here, guidance means generating feedforward reference trajectories for control systems. This paper presents terrestrial flight-test demonstrations of large diverts planned by G-FOLD onboard a vertical-takeoff/vertical-landing rocket. The G-FOLD parser, which transforms the guidance problem into a second-order cone program and so encodes the divert constraints, is described at an engineering level, including new and modified constraints incorporated for these flight tests. Several practical issues, such as discretization effects, are addressed, and the flight-test architecture is presented. A total of eight flight tests were performed. In the first three, the rocket executed diverts of increasing size preplanned by G-FOLD on the ground. Then G-FOLD was demonstrated running onboard five times. Three qualitatively different types of diverts were performed, demonstrating G-FOLD over a nontrivial area of the phase space of divert boundary conditions. With one exception not involving G-FOLD, the rocket then followed diverts planned onboard to meter-level precision, consistent with preflight performance predictions. These flight tests are believed to be the first demonstrations of large diverts with constrained, onboard powered-descent guidance, readying the way for adoption of G-FOLD by future missions.