Search

The trust-region-based sequential convex programming (TSCP) method is designed to solve the nonlinear trajectory planning problem using successive linearization and trust region. The TSCP method has a good real-time performance suitable for onboard aerospace applications. However, the main challenge in practical applications is its poor convergence. In this article, we reveal an oscillation phenomenon, which is an important factor affecting the convergence. The main contribution of this article is threefold: 1) the oscillation phenomenon is proved to be an inherent property of the TSCP method for the nonlinear trajectory planning problem; 2) the oscillation condition that determines whether oscillation occurs is obtained; and 3) the oscillation-condition-based remedy to address the oscillation is presented. The effectiveness and robustness of the proposed remedy have been verified via a complex reentry trajectory planning problem. [ABSTRACT FROM AUTHOR]

Due to the complex aerodynamic effects, the reentry trajectory optimization problem is highly nonlinear. When using sequential convex programming (SCP) methods to solve it, the iteration solution is difficult to converge. To improve this, we propose a hybrid-order soft trust region-based SCP method. We analyze the penalty effect of typical trust regions. Based on the analysis, we develop a hybrid-order soft trust region combining a small-weight first-order component and a higher-order components. To solve the subproblem reliably and effectively, we equivalently reformulate it as a second-order cone programming (SOCP) form through the relaxation technique. Combined with the line search method, we further design an SCP algorithm with guaranteed convergence under some assumptions. In the numerical simulations, the effectiveness and robustness of the proposed method have been verified using a nominal case and 200 Monte Carlo cases. [ABSTRACT FROM AUTHOR]

Sequential convex programming method is one of the potential approaches to make trajectory generation onboard. However, the contradiction between the solution accuracy and the computational efficiency restricts its application. To overcome the difficulty, a novel sequential convex programming method based on customized adaptive mesh refinement is proposed in this paper. The main contribution includes two aspects. Firstly, a customized adaptive mesh refinement method is developed. The mesh points are adjusted adaptively based on the linearization error after each iteration, and thus the number of mesh points is changed. Secondly, an unconventional convergence condition is proposed, which can make the sequential convex programming method converges to a feasible solution of the original problem with fewer iterations. Taking the entry trajectory planning problem as an example, the simulation results show that the proposed method can decrease the number of mesh points while ensuring the feasibility of converged solution. Due to the fewer iterations, the computational time to solve the original problem is decreased. As a result, a good balance between solution accuracy and computational efficiency is achieved. [ABSTRACT FROM AUTHOR]