Linear programming algorithms for worst-case identification in l/sub 1/

We formulate the uncertainty set of a worst-case identification problem using the l/sub 1/ error norm as the feasible region of a linear program. This formulation allows us to present two identification algorithms. The first one simply computes a plant in the uncertainty set. Thus, it is guaranteed a worst-case error less than or equal to the diameter of the uncertainty set. The second algorithm is more refined. It uses information about the slack variables to push the worst-case error down towards the radius of the uncertainty set, which is the optimal worst-case error.