DocumentCode :
3278945
Title :
A graph reduction for bounding the value of side information in shortest path optimization
Author :
Rinehart, M. ; Dahleh, M.A.
fYear :
2010
fDate :
June 30 2010-July 2 2010
Firstpage :
4078
Lastpage :
4083
Abstract :
Consider an agent who seeks to traverse the shortest path in a graph having random edge weights. If the agent has no side information about the realizations of the edge weights, it should simply take the path of least average length, a deterministic optimization. We consider a generalization of this framework whereby the agent has access to a limited amount of side information about the edge weights ahead of choosing a path. Specifically, we define a notion of information and information capacity, provide bounds on the agent´s performance relative to the amount of side information it receives, and offer algorithms for optimizing information within a capacity constraint. The results are based on a new graph reduction for shortest path optimization that strikes a balance between the amount of information about the graph and the distribution of the edge weights used to compute performance bounds.
Keywords :
boundary-value problems; graph theory; optimisation; software agents; agent; graph reduction side information; shortest path optimization; value bounding; Algorithm design and analysis; Constraint optimization; Distributed computing; Gaussian distribution; Performance analysis; Performance gain; Random variables; Reactive power; Stochastic processes; Topology;
fLanguage :
English
Publisher :
ieee
Conference_Titel :
American Control Conference (ACC), 2010
Conference_Location :
Baltimore, MD
ISSN :
0743-1619
Print_ISBN :
978-1-4244-7426-4
Type :
conf
DOI :
10.1109/ACC.2010.5530627
Filename :
5530627
Link To Document :
بازگشت