Computationally efficient combined design and control optimization using a coupling measure

This paper presents a novel approach to the optimization of a dynamic system's design and control. Traditionally, these problems have been solved either sequentially or in a combined manner. We propose a novel approach that uses a previously-derived coupling measure to quantify the impact of plant design variables on optimal control cost. This proposed approach has two key advantages. First, because the coupling term quantifies the gradient of the control optimization objective with respect to plant design variables, the approach ensures combined plant/control optimality. Second, because the coupling term equals the integral of optimal control co-states multiplied by static gradient terms that can be computed a priori, the proposed approach is computationally attractive. We illustrate this approach using an example cantilever beam structural design and vibration control problem. The results show significant computational cost improvements compared to traditional combined plant/control optimization. This reduction in computational cost becomes more pronounced as the number of plant design variables increases.