Model-based update in task-level feedforward control using on-line approximation (Q5930061): Difference between revisions

The dynamic models of some space systems are not perfectly accurate and may change with time. The paper proposes and studies an on-line algorithm for task-level control which improves the system operation from maneuver to maneuver by updating the time profiles of the feedforward control that are applied in each maneuver. This is done by approximating the feedforward profile dependence on the maneuver parameters using a radial basis function network, and updating this approximation after each maneuver on the base of the observed errors. Mathematically, this problem is reduced to design and convergence analysis of some optimization algorithm. The theoretical analysis is complemented by a simulation for a flexible spacecraft slewing.