Robust control of systems with input and state constraints via LQG-balancing

A novel approach to robust control of systems with input and state constraints is suggested. First, a robustification is obtained by embedding the nominal model in a whole class of extended systems. This embedding is suggested as an 'inverse' to the model reduction technique based on balanced truncation. LQG (linear quadratic Gaussian) balancing is favored because of its 'naturalness' with respect to the given constraints. Open-loop balancing is not appropriate because of its restriction to stable systems. In the final step, a saturating controller is designed for the class of nonnominal systems based on positive invariance.<>