Adaptive Continuation Strategies for Indirect Trajectory Optimization

Optimal control theory provides a mathematical foundation for solving a wide range of trajectory optimization problems. Indirect optimization methods use optimal control theory to provide high-quality solutions, but they are often regarded as inferior to other methods due to the difficulty of providing a good initial guess. Continuation addresses this difficulty by solving a series of incrementally more difficult subproblems which ultimately lead to the desired solution. The continuation process is controlled by evolving one or more parameters related to the optimal control problem. Currently, the order in which the continuation parameters are evolved must be determined by the designer. If many parameters are used or solutions are difficult to compute, this may require significant designer effort. This thesis presents strategies of automating the continuation process by drawing inspiration from methods in artificial intelligence path finding. These strategies adapt to difficult regions of the continuation space and intelligently search for valid continuation paths. In addition to increasing autonomy, adaptive continuation strategies improve the practicality of indirect methods by providing higher robustness, faster run times, and the ability to identify multiple optimal solutions