Interior-Point Algorithms for Control Allocation

Abstract

Linear-programming formulations of control allocation problems are considered, including those associated with direct allocation and mixed 1 -norm objectives. Primal-dual and predictor-corrector path-following interior-point algorithms, that are shown to be well suited for the control-allocation problems, are described in some detail with an emphasis on preferred implementations. The performance of each algorithm is evaluated for computational efficiency and for accuracy using linear models of a C-17 transport and a tailless fighter aircraft. Appropriate choices of stopping tolerances and other algorithm parameters are studied. Comparisons of speed and accuracy are made to the simplex method. Results show that real-time implementation of the algorithms is feasible, without requiring excessive number of computations.