Spacecraft actuator alignment determination through null-motion excitation

Current methods of system identification inject random or sinusoidal signals into the system and obtain feedback to learn or infer system parameters. In general, these methods do not consider the learning convergence property of the signal or its effects upon the overall system. In addition, the richness of a random signal is reduced when passed through the actuator that acts as a low-pass filter. Furthermore, the choice of both random and sinusoidal signals typically do not consider the effect of the motion on the controlling body (e.g., movement of the tool point for a robot manipulator with respect to its joint motion, movement of the entire satellite with respect to its attitude actuators, or movement of the Mars rover with respect to its wheels). A design of experiments is developed here to support reaction wheel assembly parameter identification to satisfy persistence of excitation, thereby learning system parameters without inducing large perturbations to the controllable body (e.g., spacecraft bus). This approach exploits the null-motion solutions of overactuated systems and provides a gradient-based method to identify the direction along the null space to be excited for fast local convergence of the learned parameters. The discussed method hypothesizes that a family of null-motion solutions from the excitation law will cause small but measurable perturbations upon the controllable body. Also, the motion of the actuators has the ability of being of larger amplitude and frequency than would be available structurally for the controllable body states, which makes it valid for some current satellite programs. The outcome of this research is a design of experiments structured for direct nonlinear adaptive control for an overactuated system to determine actuator alignment. This is accomplished using excitation through the null space, combined with a proposed gradient-based method (all assuming known mass and environment properties).