Abstracts and Titles for Colloquium on Predictive Control a Convex and Tractable Formulation for Robust Predictive Control Nonlinear Model Predictive Control Using Automatic Differentiation

s and titles for Collo April 4 2005 at the U A convex and tractable formulation for robu Eric Kerrigan ([email protected]) University of Cambridge The problem of finding optimal control law disturbances or model uncertainty is relative results, which allow for the explicit incorpo in the optimal control problem, are either to intractable. Much of current research in opt non-conservative and computationally effic uncertain systems with constraints on the st This talk will consider the problem of findin affine state feedback control laws that guara constraints over a finite horizon, despite the state. We will assume that the system is line constraints on the state, input and disturban well known that this control problem is non which implies that it is very difficult to find this control problem can be solved in a tract control problem and formulating it as a con outline how this alternative parameterisatio receding horizon controllers with robust sta Nonlinear Model Predictive Control using A Yi Cao ([email protected]) Cranfield University The computational burden, which obstacles techniques to be widely adopted, is mainly set of nonlinear differentiation equations an problem in real-time online. In this work, an alleviate the computational burden. The new differentiation techniques to solve the set of same time to produce the differential sensiti Page 1 Department of Automatic Control and Systems Engineering quium on Predictive Control niversity of Sheffield st predictive control s for constrained linear systems without ly well understood. However, existing ration of disturbances or model mismatch o conservative or computationally imal control is therefore aimed at finding ient methods for the optimal control of ate and input. g, at each sample instant, a sequence of ntees the satisfaction of input and state presence of a bounded disturbance on the ar and discrete-time, and that the ce are described by linear inequalities. It is -convex in the space of decision variables, a solution. However, we will show that able fashion by re-parameterising the vex optimisation problem. We will also n can be used to efficiently compute bility guarantees. utomatic Differentiation Nonlinear Model Predictive Control associated with the requirement to solve a d a nonlinear dynamic optimisation efficient algorithm has been developed to approach uses the automatic nonlinear differentiation equations, at the vity of the solution against input