Optimal thruster selection with robust estimation for formation flying applications

Abstract

The research goal was to develop a computationally fast mapper that can be easily configured to any spacecraft with various types of actuators. The estimation process must be compatible to the mapper and have a fast yet robust fault detection algorithm. A robust fault detection system must be sensitive to failures without raising any false alarms. The linear program (LP) mapping system presented in this thesis was first introduced by Crawford in 1968 (Ref. 42) and has been since used on a number of vehicle control systems (Ref. 9, Ref. 41). In this paper, several new innovations are developed as extensions to the basic LP. First, a solution scheme is presented to handle thruster performance degradation due to fuel flow loss from multiple thruster usages. Second, new techniques for solving linear programming problems with uncertain data are explored. In particular, a robust LP (RLP) formulation is developed here to deal with uncertainty in either the thruster performance or the velocity and positional measurements. The estimation process is a combination of a Kalman filter and a Generalized Likelihood Ratio (GLR) test for actuator fault detection. A new model-comparison (MC) approach is introduced in conjunction with the GLR test to quickly and reliably determine the exact nature of the failure, once a malfunction has been detected. Finally, the simulations of the estimator and thrust mapper system are performed on the SPHERES and Orion formation flying test beds and results show improved control capabilities along with significant fuel saving.