Nogood based asynchronous distributed optimization (ADOPT ng)
AAMAS '06 Proceedings of the fifth international joint conference on Autonomous agents and multiagent systems
ADOPT-ing: unifying asynchronous distributed optimization with asynchronous backtracking
Autonomous Agents and Multi-Agent Systems
| Hi-index | 0.00 |
Each agent has private problems. Private concerns can often beformulated in a general framework such as constraint satisfaction(where everything is modeled by either variables, values, orconstraints). Often agents need to find agreement with others forthe allocation of final resources. The constraint satisfaction(CSPs) is only a special case of optimization. Here it is shown howa very general technique for Distributed CSPs, Replica-basedMultiply Asynchronous Search (R-MAS) (comprising ABT, ABTR, AAS,DMAC, DMAC-ABT), can be extended and applied to optimizationproblems in distributed Weighted CSPs (WCSPs). Centralized WCSPscan be seen as MAX-CSPs where several constraints can link the samevariables. PFC-MRDAC is a good approach to MAX-CSPs. How toasynchronize and adapt it to distributed WCSPs? This articledescribes how asynchronous consistency maintenance can beintroduced in Adopt and in Asynchronous Branch & Bound. Themain new ideas proposed in this article are that: (a) a conceptcalled Weighted Consistency Nogood (WCN) allows to maintainconsistency in DisWC-SPs,(b) leading to an asynchronous equivalentof PFC-MRDAC.(c) The feedback that B&B or Adopt needs about lowbounds can be extracted from such WCNs.