When is OGY Control More Than Just Pole Placement

Deterministic chaotic systems give rise to very complex dynamical behavior in spite of the fact that they are usually of very low dimension. The exceptionally high sensitivity to changes in initial conditions makes dealing with these systems non-trivial due to inherent measurement errors and limited numerical precision. Within a chaotic attractor it is usually possible to identify unstable point attractors and unstable periodic orbits. One of the widely accepted techniques for stabilizing unstable point attractors and periodic orbits is the so called OGY controller method developed by Ott, Grebogi, and Yorke. In the case of stabilizing an unstable point attractor within the chaotic map, the OGY controller effectively forces the system to move on the corresponding stable manifold in the state-space. On the other hand, it can be shown that this strategy coincides with the design of a full state feedback controller that places the unstable eigenvalues of the state matrix of the locally linearized system at zero while keeping the location of the stable eigenvalues unchanged. When the stabilization of the original, open-loop system is the only goal, the OGY approach does not offer any advantage over other stabilizing control actions such as pole placement to arbitrary but stable eigenvalue locations. In addition, neither OGY nor arbitrary pole placement explicitly deal with the problem of noise sensitivity. Hence, when the original chaotic system is noise corrupted, it is not clear if the OGY approach is even close to being optimal with respect to noise cancelation with "acceptable" control action. Although the answer to this question is negative in the general case, there are situations where the optimal system behavior is achieved. The contribution of this paper is twofold. The first contribution is the rigorous discussion of the relationship between the OGY approach and the pole placement and the extension of the OGY method to stabilization of systems with more than one unstable eigenvalue. The second contribution of the paper is the proof that the OGY control approach to the equilibrium point stabilization in the presence of persistent, magnitude bounded process noise is actually optimal when the system performance is measured by the l∞-norm of the control signal. The proof is based on properties of the optimal l1-norm controller design and on the parameterization of all stabilizing controllers for a given linear system.