Optimal strategies for selecting from multiple asteroid interception targets

Astrodynamics;Orbital Mechanics;Optimization;Nonlinear Programming;Space;Aerospace;Optimal trajectories;Controls

De La Mata, Jorge ; Conway ; Bruce A.

关键词: Astrodynamics;    Orbital Mechanics;    Optimization;    Nonlinear Programming;    Space;    Aerospace;    Optimal trajectories;    Controls;   

Others  :  https://www.ideals.illinois.edu/bitstream/handle/2142/50647/Jorge_De%20La%20Mata.pdf?sequence=1&isAllowed=y

美国|英语

来源: The Illinois Digital Environment for Access to Learning and Scholarship

【 摘 要 】

The goal of this thesis is to demonstrate the applicability of the discrimination time optimization problem to space systems. The concept is applied to an asteroid interception problem using low-thrust propulsion and solved through a combination of particle swarm optimization (PSO) and a commercial non-linear programming (NLP) solver SNOPT (Sparse Nonlinear OPTimizer). The problem considers an intercepting vehicle and three bodies of interest, only one of which can be intercepted. The resulting optimal trajectory must maximize the amount of time the vehicle can travel before committing to intercepting one of the three bodies of interest, assuring that all three bodies are available for intercept within a maximum flight time. This time before commitment is called the discrimination time. Practically, an increase in discrimination time provides the decision makers with an opportunity for further deliberation and organization of mission priorities. This can represent additional time to determine which body holds the most scientific value, or which body presents the largest threat to human life or financial assets.The solution method is a two-step process. An approximate solution is found using a metaheursitic method, PSO. The control is modeled as a combination of polynomials to give the best opportunity for convergence to a feasible solution. The solution found by the PSO algorithm is then used to initialize a much more accurate solution method using NLP.The two optimization techniques are applied in series because of the problem complexity. The PSO algorithm is a stochastic process and can therefore not guarantee optimality and struggles to perform with a large set of optimal parameters. NLP solvers can guarantee first or second order optimality, but require initialization and convergence is contingent on the quality of the initial guess. Creating a four phase trajectory solution ad hoc that could reliably serve as an initial guess for the NLP solver is impractical, and allowing the PSO algorithm to solve a fully discrete control time history problem would be impractical. Using the two together takes advantage of the simplicity and speed provided by the PSO algorithm and the ability of the NLP solver to yield high fidelity, optimal solutions.