Dynamic Selection of Lambert Algorithms Using Performance Comparison Map

Abstract

Solving Lambert’s problem plays a critical role in the preliminary design of various space missions and ballistic missile operation applications. Many algorithms to solve Lambert’s problem has been published, but accelerating Lambert algorithms is still crucial issue, especially for solving a large number of Lambert’s problems. This dissertation proposes a dynamic algorithm selection procedure to accelerate the Lambert's problem routine composed of two phases: the preparation phase and the execution phase. In the preparation phase, a performance comparison framework for different Lambert algorithms using their characteristic parameters and cost formulae is used to generate a performance map containing the information on best algorithms for test problem instances. In the execution phase, the best algorithm for the given problem instance is selected based on the algorithm selection parameters and the performance map created in the preparation phase, and the instance is solved using the selected algorithm. The comparison results of experiments to solve a large number of test problem instances using the proposed procedure and the traditional static algorithm selection for three popular zero-revolution Lambert algorithms demonstrate the effectiveness of the dynamics algorithm selection. However, the case study of the dynamic selection with four multiple-revolution Lambert algorithm shows that the proposed dynamic selection method is not as effective as with zero-revolution Lambert algorithms. It is suggested as future work to find a novel way to choose test problem instances on which the generation of the performance comparison map is based.