Decentralized Matching Schemes for the Navy Detailing Process
Abstract
In addition to the centralized matching algorithms developed in [19], we present two decentralized matching schemes for the Navy detailing process, decentralized linear programming matching and decentralized random matching, which accommodate the requirements for matching married couples, achieve high fill rate, and fill both priority and undesirable billets. In the decentralized linear programming (DLP) matching scheme Sailors submit applications to a detailer who determines the tentative matching in each stage of a multiple-stage process until the process converges. An exact winner determination model and a heuristic are provided. A near-optimal matching is shown to be achievable. The second scheme, a decentralized random matching (DRM), allows random preference and bilateral negotiation between the Sailor and the Command. We show that it produces a stable matching, just as a centralized deferred acceptance algorithm would. A performance comparison of both centralized and decentralized matching algorithms is presented to provide guidelines on when an algorithm should be preferred and how effective each algorithm is for the Navy detailing process.
BibTeX
@techreport{Yang-2003-8828,author = {Wei Yang and Katia Sycara and Joseph Andrew Giampapa},
title = {Decentralized Matching Schemes for the Navy Detailing Process},
year = {2003},
month = {December},
institute = {Carnegie Mellon University},
address = {Pittsburgh, PA},
number = {CMU-RI-TR-03-51},
keywords = {decentralized linear programming matching, decentralized random matching, Navy detailing process, stable matching, couples},
}