Dynamic programming applied to electromagnetic satellite actuation

Abstract

Electromagnetic formation flight (EMFF) is an enabling technology for a number of space mission architectures. While much work has been done for EMFF control for large separation distances, little work has been done for close-proximity EMFF control, where the system dynamics are quite complex. Dynamic programming has been heavily used in the optimization world, but not on embedded systems. In this thesis, dynamic programming is applied to satellite control, using close-proximity EMFF control as a case study. The concepts of dynamic programming and approximate dynamic programming are discussed. Several versions of the close-proximity EMFF control problem are formulated as a dynamic programming problems. One of the formulations is used as a case study for developing and examining the cost-to-go. Methods for implementing an approximate dynamic programming controller on a satellite are discussed. Methods for resolving physical states and dynamic programming states are presented. Because the success of dynamic programming depends on the system model, a novel method for finding the mass properties of a satellite, which would likely be used in the dynamic programming model, is introduced. This method is used to characterize the mass properties of three satellite systems: SPHERES, VERTIGO, and RINGS. Finally, a method for position and attitude estimation for systems that use line-of-sight measurements that does not require the use of a model is developed. This method is useful for model validation of the models used in the dynamic programming formulation.