DisLRPɑ: ɑ-approximation in generalized mutual assignment

Authors:
Katsutoshi Hirayama
Affiliations:
Kobe University, Kobe, Japan
Venue:
Proceedings of the 6th international joint conference on Autonomous agents and multiagent systems
Year:
2007

Citing 3
Cited 0

The distributed breakout algorithms

Artificial Intelligence - Special issue: Distributed constraint satisfaction
The Lagrangian Relaxation Method for Solving Integer Programming Problems

Management Science
A new approach to distributed task assignment using Lagrangian decomposition and distributed constraint satisfaction

AAAI'06 Proceedings of the 21st national conference on Artificial intelligence - Volume 1

Quantified Score

Hi-index	0.00

Visualization

Abstract

This paper presents a new distributed solution protocol, called DisLRPα, for the Generalized Mutual Assignment Problem (GMAP). The GMAP is a typical distributed combinatorial optimization problem whose goal is to maximize social welfare of the agents. Unlike the previous protocol for the GMAP, DisLRPα can provide a theoretical guarantee on global solution quality. In DisLRPα, as with in the previous protocol, the agents repeatedly solve their local problems while coordinating their local solutions using a distributed constraint satisfaction technique. The key difference is that, in DisLRPα, each agent is required to produce a feasible solution whose local objective value is not lower than α (0 α can certainly find a solution whose global objective value is higher than that theoretically guaranteed. Furthermore, they also show that, while spending extra communication and computation costs, DisLRPα can produce a significantly better solution than the previous protocol if we set α appropriately.