Linear-quadratic worst-case control

This paper gives an applied view of linear-quadratic worst-case control and relates it to linear-quadratic-Gaussian smoothing. It extends the worst-case control problem formulation to include tracking of desired output histories and nonzero terminal constraints and separates the problem into future,past, and present problems, each of which must satisfy a conjugate-point condition. It includes a finite time horizon example, namely, a helicopter hover-position change in which the disturbance is horizontal wind velocity. Linear-quadratic best-case controllers can be obtained by using positive instead of negative weights in the quadratic performance index.