Multiple solutions to the l/sup 1/-optimal control problem and its dual linear programming problem

Abstract

This paper explores the consequences for the l/sup 1/-optimal controller of the dual linear programming problem having multiple solutions, for linear time-invariant single-input/single-output systems. When the dual problem has multiple solutions, all solutions yield the same set of optimal controllers. If these multiple solutions comprise an entire face of the constraint region, there is a single optimal controller. Thus, if the constraint region is two-dimensional, the primal and dual problems cannot both have multiple solutions. An example is given with a three dimensional constraint region where both problems have multiple solutions.