We consider the problem of designing distributed output-feedback controllers that achieve H2 and H∞ performance objectives for a group of plants dynamically interconnected via an arbitrary directed communication network. For a particular class of discrete-time linear time-invariant interconnected systems that are characterized by a structural property of their state-space matrices, we design stabilizing distributed controllers which can use the available network along with the plants of the interconnected system. This is achieved by means of parametrization for the output-feedback linear controllers that linearizes the closed-loop H2 and H∞ norm conditions and provides equivalent linear matrix inequalities (LMIs). Using these LMIs, we formulate the minimization of H 2 and H∞ norms as semi-definite programs (SDPs) that can be efficiently solved using well-established and standard techniques and tools. The solutions of these SDPs allow us to synthesize the corresponding controllers that are realizable over the given network. Even though we provide only sufficiency conditions for the design of stabilizing distributed controllers, simulations show that the synthesized controllers we obtain provide good performance in spite of being suboptimal compared to the centralized controller. In essence, we gain the advantage of designing realizable distributed controllers at the expense of some performance degradation compared to the centralized solutions.
|Original language||English (US)|
|Title of host publication||2nd IFAC Workshop on Distributed Estimation and Control in Networked Systems, NecSys'10|
|Number of pages||6|
|State||Published - 2010|
|Event||2nd IFAC Workshop on Distributed Estimation and Control in Networked Systems, NecSys'10 - Annecy, France|
Duration: Sep 13 2010 → Sep 14 2010
|Name||IFAC Proceedings Volumes (IFAC-PapersOnline)|
|Other||2nd IFAC Workshop on Distributed Estimation and Control in Networked Systems, NecSys'10|
|Period||9/13/10 → 9/14/10|
Bibliographical noteFunding Information:
This research has been supported by NSF grant ECCS0901846.
- Distributed control
- H and H optimization
- Linear matrix inequalities