Evaluation of Optimization Methods for Control Allocation
Department of Electrical & Computer Engineering
University of Utah
Control allocation has been defined in the context of flight control as the problem of distributing control requirements among multiple, redundant actuators. Position and rate limits on the control variables make the problem difficult. Traditionally, control allocation has been performed using ganging (combining different control variables through linear relationships), or using approximate solutions based on the least-squares algorithm. More recently, constrained optimization methods have been considered. In the talk, we will discuss various methods that have been proposed, and evaluate their relative benefits. Three algorithms with low computational requirements will be reviewed: a redistributed pseudo-inverse technique, a quadratic programming algorithm, and a fixed-point method. Constrained optimization algorithms obtained by transforming the control allocation problems into linear programs will also be presented. The results of the testing of the various algorithms in flight control problems will be discussed. A main conclusion of the research is that optimization methods are implement able in real-time and result in noticeable improvements of performance over simpler approaches.
Friday, September 21, 2001
3:30 - 5:00 p.m.