Structured Computation of Constrained Quadratic Optimal Controls for Heterogeneous Networks*

The problem of computing finite-horizon constrained quadratic optimal controls is considered for networks of interconnected heterogeneous linear systems. A projection based iterative algorithm is studied. An iteration of the algorithm involves n calculations, each associated with the update of the control input sequence for one of the n sub-systems in the network. In general, the update calculation for each sub-system involves information from beyond those directly connect to it. However, for particular interconnection structures only local information is required, and various examples of such structures are identified in the case of coupling via the controls or via the states. The approach is demonstrated numerically with comparison to another method available in the literature.