14.07.2013 Views

X - UWSpace - University of Waterloo

X - UWSpace - University of Waterloo

X - UWSpace - University of Waterloo

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

Chapter 6 Conclusions and Future Research<br />

6.1 Conclusions<br />

In this thesis, the main objective is to develop a new parallel decomposition method for<br />

rnulti-part linear prograrnmjng models. We focus on the pro<strong>of</strong>s and demonsuation <strong>of</strong> the<br />

convergence <strong>of</strong> the two-part and multi-part dgorithms. We repori in some test results that the<br />

algonthms converge to an optimal solution <strong>of</strong> the original problems in a finite number <strong>of</strong><br />

decomposition i terations.<br />

The idea <strong>of</strong> this thesis originally came frorn Lan [1993] for multi-stage nested decomposition.<br />

however due to some infesibility problems on applying the nested decomposition in parallel, we<br />

developed another parallel decomposition scheme for multi-part stxuctured models.<br />

The parallel decomposition algorithm for two-part models consistently converged to an<br />

optimal solution (and a convergence pro<strong>of</strong> was formed). but our fint aigorithm for multi-pan<br />

models did not always converge (we cal1 this algrithm a heuristic). Finally, we came up with a<br />

convergent pûrallel primal-dual decomposi tion aleorithm for multi-part LPs. w hich convergeci<br />

in al1 tests. by applying the two-part decomposition principle ~cursively in a hiemhical way.<br />

The distinctions between Lads algorithm and the new parallel decomposition aigorithm<br />

;ire summarized in Table 6.1.

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!