Enhanced Successive Convexification Based on Error-Feedback Index and Line Search Filter

Abstract

This paper introduces convexification error as an error-feedback optimization index into successive convexification optimization, widely used in aerospace engineering applications, such as reusable rocket landing. The error-feedback optimization index is not only a measure of the solution accuracy but also a criterion of convergence. The original single-objective successive convexification optimization problem is transformed into a bi-objective optimization problem, which optimizes both the convexification error and the objective. An efficient and robust convergence strategy is proposed based on a line search filter for the newly obtained optimization problem. The proposed convergence strategy improves the accuracy of the obtained solution and alleviates the potential sensitivity of the successive convexification optimization convergence to the initial guess. The proposed convergence strategy is verified using a universal controllability set approximation method. The numerical experimental results show that the proposed algorithms have good practical potential in aerospace engineering.