Adaptive control of decentralized systems: Known subsystems, unknown interconnections

One of the common assumptions in control of large interconnected systems is that models of subsystems are to a large extent known to a designer, and an essential modeling uncertainty resides in the interconnections. Current decentralized adaptive control schemes take no advantage of this fact. In this paper an algorithm is presented in which adaptation of local feedback gains is in the a direction which compensates for the unknown interconnections, while exploiting the knowledge about the sub-systems. As a result, we broaden considerably the class of interconnected systems for which decentralized adaptation is feasible.