Sequential convex programming without penalty function for reentry trajectory optimization problem

Abstract

Sequential convex programming (SCP) has been extensively utilized in reentry trajectory optimization due to its high computational efficiency. However, the current SCP approaches primarily rely on penalty function, where the selection of the penalty function weight presents a significant challenge. In this paper, an improved trust region shrinking SCP algorithm is proposed that separates the treatment of the objective function and constraint violation without the need for selecting penalty function weight and introduction of slack variables. Firstly, from the perspective of multi-objective optimization, the filter and acceptance condition are introduced to ensure that the proposed algorithm converges to feasible solutions and then to the optimal solution based on switching condition and sufficient condition. Then an effective feasibility restoration phase is proposed to address infeasibility of subproblems without introducing slack variables, while ensuring the robustness of the proposed algorithm. Additionally, a theoretical analysis is provided to guarantee the convergence of the algorithm. Finally, simulations are conducted to verify that the proposed algorithm demonstrates a 69.54% improvement in average solution time and stronger robustness compared to basic trust region shrinking SCP algorithm. Simultaneously, the proposed algorithm also demonstrates an advantage in solving speed compared to a particular advanced penalty function-based SCP algorithm.