Abstraction Predictive Control for Chaotic Spacecraft Orbit Design*

We formulate a predictive control algorithm for planning spacecraft trajectories at close proximity to asteroids, where state-space dynamics are sensitive, complex, and non-periodic, thus invalidating standard orbit design methods. Sets of scientific observation objectives are pursued by defining an abstracted goal space and a control law for regulating the motion in this space toward a state signifying mission completion. Non-deterministic components of the control scheme, which obtains suboptimal solutions via heuristic search of the complex control space, are found to produce diverse solution paths through the state space and goal space alike, implying high sensitivity of the approach despite its ultimate stability. The degree of this diversity is charted as a function of chaoticity in the state-space dynamics and the complexity of the goal-space definition. These results are compared to those obtained under Keplerian motion and are discussed in the context of eventual application of the algorithm for preliminary mission design and onboard implementation for robust online planning.