| Decentralization poses a significant challenge in design for the control engineer. This can be further exacerbated by uncertainty in the systems to be controlled. The interplay between these two factors warrants greater scrutiny, if we are to cope with the increasingly complex and decentralized systems arising in practice today. Hence the need for a thorough understanding of the theoretical questions faced in decentralized adaptive control.; This thesis addresses the problem of decentralized adaptive control from two main vantage points: the strictly decentralized adaptive control; and the decentralized adaptive control problem with communication amongst the controllers.; The first problem posits an ideal scenario in which no communication occurs amongst the various controllers of the decentralized system. The analysis is then concerned with the best performance that can be obtained under these circumstances. It is shown that for a large class of systems, zero error asymptotic tracking can still be attained, if the controllers have prior information in the form of the desired goals (trajectories) of the other subsystems.; The second problem addresses the problem of how communication amongst the various controllers in a decentralized systems, might be used effectively to enhance the performance of the control system. Different protocols for utilizing this information are suggested and analysis as well as simulations provided in the thesis. It is shown that these protocols lead to improvement in performance of the decentralized adaptive control system. |
